vkernel{,64} - CPU topology support
[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.
da673940
JG
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>
bf7c5253 53#include <vm/vm_map.h>
e130cdcb 54#include <sys/mplock2.h>
da673940
JG
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;
39d69dae 85long 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 */
9bea6114
MC
115int vkernel_b_arg; /* -b argument - no of logical CPU bits - only SMP */
116int vkernel_B_arg; /* -B argument - no of core bits - only SMP */
da673940
JG
117
118struct privatespace *CPU_prvspace;
119
120static struct trapframe proc0_tf;
121static void *proc0paddr;
122
123static void init_sys_memory(char *imageFile);
124static void init_kern_memory(void);
125static void init_globaldata(void);
126static void init_vkernel(void);
127static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
128static void init_netif(char *netifExp[], int netifFileNum);
278c4e67
SG
129static void writepid(void);
130static void cleanpid(void);
da673940 131static int unix_connect(const char *path);
278c4e67
SG
132static void usage_err(const char *ctl, ...);
133static void usage_help(_Bool);
da673940
JG
134
135static int save_ac;
136static char **save_av;
137
138/*
139 * Kernel startup for virtual kernels - standard main()
140 */
141int
142main(int ac, char **av)
143{
144 char *memImageFile = NULL;
145 char *netifFile[VKNETIF_MAX];
146 char *diskFile[VKDISK_MAX];
147 char *cdFile[VKDISK_MAX];
148 char *suffix;
149 char *endp;
974f8ffe 150 char *tmp;
9bea6114 151 char *tok;
da673940
JG
152 int netifFileNum = 0;
153 int diskFileNum = 0;
154 int cdFileNum = 0;
155 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
156 int c;
157 int i;
158 int j;
159 int n;
160 int isq;
974f8ffe
AHJ
161 int pos;
162 int eflag;
da673940
JG
163 int real_vkernel_enable;
164 int supports_sse;
165 size_t vsize;
974f8ffe
AHJ
166 size_t kenv_size;
167 size_t kenv_size2;
da673940
JG
168
169 save_ac = ac;
170 save_av = av;
974f8ffe
AHJ
171 eflag = 0;
172 pos = 0;
173 kenv_size = 0;
da673940
JG
174
175 /*
176 * Process options
177 */
178 kernel_mem_readonly = 1;
179#ifdef SMP
180 optcpus = 2;
9bea6114
MC
181 vkernel_b_arg = 0;
182 vkernel_B_arg = 0;
da673940
JG
183#endif
184 lwp_cpu_lock = LCL_NONE;
185
186 real_vkernel_enable = 0;
187 vsize = sizeof(real_vkernel_enable);
188 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
189
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 }
194
195 real_ncpus = 1;
196 vsize = sizeof(real_ncpus);
197 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
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) {
da673940
JG
203 switch(c) {
204 case 'e':
205 /*
206 * name=value:name=value:name=value...
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.
212 */
213 n = strlen(optarg);
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) {
da673940
JG
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;
237 else
238 kern_envp[j++] = optarg[i];
239 }
240 kern_envp[j++] = 0;
241 kern_envp[j++] = 0;
974f8ffe 242 eflag++;
da673940
JG
243 break;
244 case 's':
245 boothowto |= RB_SINGLE;
246 break;
247 case 'v':
248 bootverbose = 1;
249 break;
250 case 'i':
251 memImageFile = optarg;
252 break;
253 case 'I':
254 if (netifFileNum < VKNETIF_MAX)
255 netifFile[netifFileNum++] = strdup(optarg);
256 break;
257 case 'r':
258 if (bootOnDisk < 0)
259 bootOnDisk = 1;
260 if (diskFileNum + cdFileNum < VKDISK_MAX)
261 diskFile[diskFileNum++] = strdup(optarg);
262 break;
263 case 'c':
264 if (bootOnDisk < 0)
265 bootOnDisk = 0;
266 if (diskFileNum + cdFileNum < VKDISK_MAX)
267 cdFile[cdFileNum++] = strdup(optarg);
268 break;
269 case 'm':
270 Maxmem_bytes = strtoull(optarg, &suffix, 0);
271 if (suffix) {
272 switch(*suffix) {
273 case 'g':
274 case 'G':
275 Maxmem_bytes <<= 30;
276 break;
277 case 'm':
278 case 'M':
279 Maxmem_bytes <<= 20;
280 break;
281 case 'k':
282 case 'K':
283 Maxmem_bytes <<= 10;
284 break;
285 default:
286 Maxmem_bytes = 0;
278c4e67 287 usage_err("Bad maxmem option");
da673940
JG
288 /* NOT REACHED */
289 break;
290 }
291 }
292 break;
293 case 'l':
294 next_cpu = -1;
295 if (strncmp("map", optarg, 3) == 0) {
296 lwp_cpu_lock = LCL_PER_CPU;
297 if (optarg[3] == ',') {
298 next_cpu = strtol(optarg+4, &endp, 0);
299 if (*endp != '\0')
278c4e67 300 usage_err("Bad target CPU number at '%s'", endp);
da673940
JG
301 } else {
302 next_cpu = 0;
303 }
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);
da673940
JG
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);
da673940 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);
da673940
JG
315 }
316 break;
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, ":");
da673940 323#ifdef SMP
9bea6114 324 optcpus = strtol(tok, NULL, 0);
da673940 325 if (optcpus < 1 || optcpus > MAXCPU)
278c4e67 326 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
9bea6114
MC
327
328 /* :lbits argument */
329 tok = strtok(NULL, ":");
330 if (tok != NULL) {
331 vkernel_b_arg = strtol(tok, NULL, 0);
332
333 /* :cbits argument */
334 tok = strtok(NULL, ":");
335 if (tok != NULL) {
336 vkernel_B_arg = strtol(tok, NULL, 0);
337 }
338
339 }
340
da673940 341#else
9bea6114 342 if (strtol(tok, NULL, 0) != 1) {
278c4e67 343 usage_err("You built a UP vkernel, only 1 cpu!");
da673940 344 }
9bea6114
MC
345
346 /* :lbits argument */
347 tok = strtok(NULL, ":");
348 if (tok != NULL) {
349 usage_err("You built a UP vkernel. No CPU topology available");
350
351 /* :cbits argument */
352 tok = strtok(NULL, ":");
353 if (tok != NULL) {
354 usage_err("You built a UP vkernel. No CPU topology available");
355 }
356 }
da673940
JG
357#endif
358
9bea6114 359
da673940
JG
360 case 'p':
361 pid_file = optarg;
362 break;
363 case 'U':
364 kernel_mem_readonly = 0;
365 break;
278c4e67
SG
366 case 'h':
367 usage_help(true);
368 break;
278c4e67
SG
369 default:
370 usage_help(false);
da673940
JG
371 }
372 }
373
374 writepid();
375 cpu_disable_intr();
376 init_sys_memory(memImageFile);
377 init_kern_memory();
378 init_globaldata();
379 init_vkernel();
380 setrealcpu();
381 init_kqueue();
382
08771751
VS
383 vmm_guest = 1;
384
da673940
JG
385 /*
386 * Check TSC
387 */
388 vsize = sizeof(tsc_present);
389 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
390 vsize = sizeof(tsc_frequency);
391 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
392 if (tsc_present)
393 cpu_feature |= CPUID_TSC;
394
395 /*
396 * Check SSE
397 */
398 vsize = sizeof(supports_sse);
399 supports_sse = 0;
400 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
401 init_fpu(supports_sse);
402 if (supports_sse)
403 cpu_feature |= CPUID_SSE | CPUID_FXSR;
404
405 /*
406 * We boot from the first installed disk.
407 */
408 if (bootOnDisk == 1) {
409 init_disk(diskFile, diskFileNum, VKD_DISK);
410 init_disk(cdFile, cdFileNum, VKD_CD);
411 } else {
412 init_disk(cdFile, cdFileNum, VKD_CD);
413 init_disk(diskFile, diskFileNum, VKD_DISK);
414 }
415 init_netif(netifFile, netifFileNum);
416 init_exceptions();
417 mi_startup();
418 /* NOT REACHED */
278c4e67 419 exit(EX_SOFTWARE);
da673940
JG
420}
421
422/*
423 * Initialize system memory. This is the virtual kernel's 'RAM'.
424 */
425static
426void
427init_sys_memory(char *imageFile)
428{
429 struct stat st;
430 int i;
431 int fd;
432
433 /*
434 * Figure out the system memory image size. If an image file was
435 * specified and -m was not specified, use the image file's size.
436 */
da673940
JG
437 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
438 Maxmem_bytes = (vm_paddr_t)st.st_size;
439 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
440 Maxmem_bytes == 0) {
376264e3 441 errx(1, "Cannot create new memory file %s unless "
da673940
JG
442 "system memory size is specified with -m",
443 imageFile);
444 /* NOT REACHED */
445 }
446
447 /*
448 * Maxmem must be known at this time
449 */
242e00e6
AHJ
450 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
451 errx(1, "Bad maxmem specification: 64MB minimum, "
da673940
JG
452 "multiples of %dMB only",
453 SEG_SIZE / 1024 / 1024);
454 /* NOT REACHED */
455 }
456
457 /*
458 * Generate an image file name if necessary, then open/create the
459 * file exclusively locked. Do not allow multiple virtual kernels
460 * to use the same image file.
bc3cc25e
MD
461 *
462 * Don't iterate through a million files if we do not have write
463 * access to the directory, stop if our open() failed on a
464 * non-existant file. Otherwise opens can fail for any number
da673940
JG
465 */
466 if (imageFile == NULL) {
467 for (i = 0; i < 1000000; ++i) {
468 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
469 fd = open(imageFile,
470 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
bc3cc25e 471 if (fd < 0 && stat(imageFile, &st) == 0) {
da673940
JG
472 free(imageFile);
473 continue;
474 }
475 break;
476 }
477 } else {
478 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
479 }
480 fprintf(stderr, "Using memory file: %s\n", imageFile);
481 if (fd < 0 || fstat(fd, &st) < 0) {
482 err(1, "Unable to open/create %s", imageFile);
483 /* NOT REACHED */
484 }
485
486 /*
b72e51d1
MD
487 * Truncate or extend the file as necessary. Clean out the contents
488 * of the file, we want it to be full of holes so we don't waste
489 * time reading in data from an old file that we no longer care
490 * about.
da673940 491 */
b72e51d1
MD
492 ftruncate(fd, 0);
493 ftruncate(fd, Maxmem_bytes);
da673940 494
da673940
JG
495 MemImageFd = fd;
496 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
9f90716f 497 physmem = Maxmem;
da673940
JG
498}
499
500/*
501 * Initialize kernel memory. This reserves kernel virtual memory by using
502 * MAP_VPAGETABLE
503 */
504
505static
506void
507init_kern_memory(void)
508{
509 void *base;
510 void *try;
da673940
JG
511 char dummy;
512 char *topofstack = &dummy;
da673940
JG
513 int i;
514 void *firstfree;
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.
520 *
521 * The memory map must be segment-aligned so we can properly
522 * offset KernelPTD.
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.
527 */
528 try = (void *)(512UL << 30);
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,
2c2e847c 533 MemImageFd, (off_t)try);
da673940
JG
534 if (base == try)
535 break;
536 if (base != MAP_FAILED)
537 munmap(base, KERNEL_KVA_SIZE);
538 try = (char *)try + (512UL << 30);
539 }
540 if (base != try) {
541 err(1, "Unable to mmap() kernel virtual memory!");
542 /* NOT REACHED */
543 }
544 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
545 KvaStart = (vm_offset_t)base;
546 KvaSize = KERNEL_KVA_SIZE;
547 KvaEnd = KvaStart + KvaSize;
d4f5f27c
MD
548
549 /* cannot use kprintf yet */
da673940
JG
550 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
551
552 /* MAP_FILE? */
553 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
554 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
555 MemImageFd, 0);
556 if (dmap_min_address == MAP_FAILED) {
557 err(1, "Unable to mmap() kernel DMAP region!");
558 /* NOT REACHED */
559 }
560
4090d6ff 561 firstfree = NULL;
2c2e847c 562 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
da673940
JG
563
564 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
565 0 | VPTE_R | VPTE_W | VPTE_V);
566
567 /*
568 * phys_avail[] represents unallocated physical memory. MI code
569 * will use phys_avail[] to create the vm_page array.
570 */
2c2e847c 571 phys_avail[0] = (vm_paddr_t)firstfree;
da673940
JG
572 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
573 phys_avail[1] = Maxmem_bytes;
574
575#if JGV
576 /*
577 * (virtual_start, virtual_end) represent unallocated kernel virtual
578 * memory. MI code will create kernel_map using these parameters.
579 */
580 virtual_start = KvaStart + (long)firstfree;
581 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
582 virtual_end = KvaStart + KERNEL_KVA_SIZE;
583#endif
584
585 /*
586 * pmap_growkernel() will set the correct value.
587 */
588 kernel_vm_end = 0;
589
590 /*
591 * Allocate space for process 0's UAREA.
592 */
593 proc0paddr = (void *)virtual_start;
594 for (i = 0; i < UPAGES; ++i) {
595 pmap_kenter_quick(virtual_start, phys_avail[0]);
596 virtual_start += PAGE_SIZE;
597 phys_avail[0] += PAGE_SIZE;
598 }
599
600 /*
601 * crashdumpmap
602 */
603 crashdumpmap = virtual_start;
604 virtual_start += MAXDUMPPGS * PAGE_SIZE;
605
606 /*
607 * msgbufp maps the system message buffer
608 */
609 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
610 msgbufp = (void *)virtual_start;
611 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
612 pmap_kenter_quick(virtual_start, phys_avail[0]);
613 virtual_start += PAGE_SIZE;
614 phys_avail[0] += PAGE_SIZE;
615 }
616 msgbufinit(msgbufp, MSGBUF_SIZE);
617
618 /*
619 * used by kern_memio for /dev/mem access
620 */
621 ptvmmap = (caddr_t)virtual_start;
622 virtual_start += PAGE_SIZE;
623
624 /*
625 * Bootstrap the kernel_pmap
626 */
627#if JGV
628 pmap_bootstrap();
629#endif
630}
631
632/*
633 * Map the per-cpu globaldata for cpu #0. Allocate the space using
634 * virtual_start and phys_avail[0]
635 */
636static
637void
638init_globaldata(void)
639{
640 int i;
641 vm_paddr_t pa;
642 vm_offset_t va;
643
644 /*
645 * Reserve enough KVA to cover possible cpus. This is a considerable
646 * amount of KVA since the privatespace structure includes two
647 * whole page table mappings.
648 */
649 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
650 CPU_prvspace = (void *)virtual_start;
651 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
652
653 /*
654 * Allocate enough physical memory to cover the mdglobaldata
655 * portion of the space and the idle stack and map the pages
656 * into KVA. For cpu #0 only.
657 */
658 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
659 pa = phys_avail[0];
660 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
661 pmap_kenter_quick(va, pa);
662 phys_avail[0] += PAGE_SIZE;
663 }
664 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
665 pa = phys_avail[0];
666 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
667 pmap_kenter_quick(va, pa);
668 phys_avail[0] += PAGE_SIZE;
669 }
670
671 /*
672 * Setup the %gs for cpu #0. The mycpu macro works after this
673 * point. Note that %fs is used by pthreads.
674 */
675 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
676}
677
678/*
679 * Initialize very low level systems including thread0, proc0, etc.
680 */
681static
682void
683init_vkernel(void)
684{
685 struct mdglobaldata *gd;
686
687 gd = &CPU_prvspace[0].mdglobaldata;
688 bzero(gd, sizeof(*gd));
689
690 gd->mi.gd_curthread = &thread0;
691 thread0.td_gd = &gd->mi;
692 ncpus = 1;
693 ncpus2 = 1; /* rounded down power of 2 */
694 ncpus_fit = 1; /* rounded up power of 2 */
695 /* ncpus2_mask and ncpus_fit_mask are 0 */
696 init_param1();
697 gd->mi.gd_prvspace = &CPU_prvspace[0];
698 mi_gdinit(&gd->mi, 0);
699 cpu_gdinit(gd, 0);
700 mi_proc0init(&gd->mi, proc0paddr);
701 lwp0.lwp_md.md_regs = &proc0_tf;
702
703 /*init_locks();*/
e130cdcb
MD
704#ifdef SMP
705 /*
706 * Get the initial mplock with a count of 1 for the BSP.
707 * This uses a LOGICAL cpu ID, ie BSP == 0.
708 */
709 cpu_get_initial_mplock();
710#endif
da673940
JG
711 cninit();
712 rand_initialize();
713#if 0 /* #ifdef DDB */
714 kdb_init();
715 if (boothowto & RB_KDB)
716 Debugger("Boot flags requested debugger");
717#endif
718 identcpu();
719#if 0
720 initializecpu(); /* Initialize CPU registers */
721#endif
722 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
723
724#if 0
725 /*
726 * Map the message buffer
727 */
728 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
729 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
730 msgbufinit(msgbufp, MSGBUF_SIZE);
731#endif
732#if 0
733 thread0.td_pcb_cr3 ... MMU
734 lwp0.lwp_md.md_regs = &proc0_tf;
735#endif
736}
737
738/*
739 * Filesystem image paths for the virtual kernel are optional.
740 * If specified they each should point to a disk image,
741 * the first of which will become the root disk.
742 *
743 * The virtual kernel caches data from our 'disk' just like a normal kernel,
744 * so we do not really want the real kernel to cache the data too. Use
745 * O_DIRECT to remove the duplication.
746 */
747static
748void
749init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
750{
751 int i;
752
753 if (diskFileNum == 0)
754 return;
755
756 for(i=0; i < diskFileNum; i++){
757 char *fname;
758 fname = diskExp[i];
759
760 if (fname == NULL) {
761 warnx("Invalid argument to '-r'");
762 continue;
763 }
764
765 if (DiskNum < VKDISK_MAX) {
766 struct stat st;
767 struct vkdisk_info* info = NULL;
768 int fd;
769 size_t l = 0;
770
771 if (type == VKD_DISK)
6882b7a5 772 fd = open(fname, O_RDWR|O_DIRECT, 0644);
da673940
JG
773 else
774 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
775 if (fd < 0 || fstat(fd, &st) < 0) {
da673940
JG
776 err(1, "Unable to open/create %s", fname);
777 /* NOT REACHED */
778 }
6882b7a5
MD
779 if (S_ISREG(st.st_mode)) {
780 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
781 errx(1, "Disk image %s is already "
782 "in use\n", fname);
783 /* NOT REACHED */
784 }
785 }
da673940
JG
786
787 info = &DiskInfo[DiskNum];
788 l = strlen(fname);
789
790 info->unit = i;
791 info->fd = fd;
792 info->type = type;
793 memcpy(info->fname, fname, l);
794
795 if (DiskNum == 0) {
796 if (type == VKD_CD) {
797 rootdevnames[0] = "cd9660:vcd0a";
798 } else if (type == VKD_DISK) {
799 rootdevnames[0] = "ufs:vkd0s0a";
800 rootdevnames[1] = "ufs:vkd0s1a";
801 }
802 }
803
804 DiskNum++;
805 } else {
806 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
807 continue;
808 }
809 }
810}
811
812static
813int
814netif_set_tapflags(int tap_unit, int f, int s)
815{
816 struct ifreq ifr;
817 int flags;
818
819 bzero(&ifr, sizeof(ifr));
820
821 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
822 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
823 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
824 return -1;
825 }
826
827 /*
828 * Adjust if_flags
829 *
830 * If the flags are already set/cleared, then we return
831 * immediately to avoid extra syscalls
832 */
833 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
834 if (f < 0) {
835 /* Turn off flags */
836 f = -f;
837 if ((flags & f) == 0)
838 return 0;
839 flags &= ~f;
840 } else {
841 /* Turn on flags */
842 if (flags & f)
843 return 0;
844 flags |= f;
845 }
846
847 /*
848 * Fix up ifreq.ifr_name, since it may be trashed
849 * in previous ioctl(SIOCGIFFLAGS)
850 */
851 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
852
853 ifr.ifr_flags = flags & 0xffff;
854 ifr.ifr_flagshigh = flags >> 16;
855 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
856 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
857 return -1;
858 }
859 return 0;
860}
861
862static
863int
864netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
865{
866 struct ifaliasreq ifra;
867 struct sockaddr_in *in;
868
869 bzero(&ifra, sizeof(ifra));
870 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
871
872 /* Setup address */
873 in = (struct sockaddr_in *)&ifra.ifra_addr;
874 in->sin_family = AF_INET;
875 in->sin_len = sizeof(*in);
876 in->sin_addr.s_addr = addr;
877
878 if (mask != 0) {
879 /* Setup netmask */
880 in = (struct sockaddr_in *)&ifra.ifra_mask;
881 in->sin_len = sizeof(*in);
882 in->sin_addr.s_addr = mask;
883 }
884
885 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
886 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
887 return -1;
888 }
889 return 0;
890}
891
892static
893int
894netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
895{
896 struct ifbreq ifbr;
897 struct ifdrv ifd;
898
899 bzero(&ifbr, sizeof(ifbr));
900 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
901 "tap%d", tap_unit);
902
903 bzero(&ifd, sizeof(ifd));
904 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
905 ifd.ifd_cmd = BRDGADD;
906 ifd.ifd_len = sizeof(ifbr);
907 ifd.ifd_data = &ifbr;
908
909 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
910 /*
911 * 'errno == EEXIST' means that the tap(4) is already
912 * a member of the bridge(4)
913 */
914 if (errno != EEXIST) {
915 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
916 return -1;
917 }
918 }
919 return 0;
920}
921
922#define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
923
924/*
925 * Locate the first unused tap(4) device file if auto mode is requested,
926 * or open the user supplied device file, and bring up the corresponding
927 * tap(4) interface.
928 *
929 * NOTE: Only tap(4) device file is supported currently
930 */
931static
932int
933netif_open_tap(const char *netif, int *tap_unit, int s)
934{
935 char tap_dev[MAXPATHLEN];
936 int tap_fd, failed;
937 struct stat st;
938 char *dname;
939
940 *tap_unit = -1;
941
942 if (strcmp(netif, "auto") == 0) {
943 /*
944 * Find first unused tap(4) device file
945 */
946 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
947 if (tap_fd < 0) {
948 warnc(errno, "Unable to find a free tap(4)");
949 return -1;
950 }
951 } else {
952 /*
953 * User supplied tap(4) device file or unix socket.
954 */
955 if (netif[0] == '/') /* Absolute path */
956 strlcpy(tap_dev, netif, sizeof(tap_dev));
957 else
958 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
959
960 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
961
962 /*
963 * If we cannot open normally try to connect to it.
964 */
965 if (tap_fd < 0)
966 tap_fd = unix_connect(tap_dev);
967
968 if (tap_fd < 0) {
969 warn("Unable to open %s", tap_dev);
970 return -1;
971 }
972 }
973
974 /*
975 * Check whether the device file is a tap(4)
976 */
977 if (fstat(tap_fd, &st) < 0) {
978 failed = 1;
979 } else if (S_ISCHR(st.st_mode)) {
980 dname = fdevname(tap_fd);
981 if (dname)
982 dname = strstr(dname, "tap");
983 if (dname) {
984 /*
985 * Bring up the corresponding tap(4) interface
986 */
987 *tap_unit = strtol(dname + 3, NULL, 10);
988 printf("TAP UNIT %d\n", *tap_unit);
989 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
990 failed = 0;
3b6e5cb0
SW
991 else
992 failed = 1;
da673940
JG
993 } else {
994 failed = 1;
995 }
996 } else if (S_ISSOCK(st.st_mode)) {
997 /*
998 * Special socket connection (typically to vknet). We
999 * do not have to do anything.
1000 */
1001 failed = 0;
1002 } else {
1003 failed = 1;
1004 }
1005
1006 if (failed) {
1007 warnx("%s is not a tap(4) device or socket", tap_dev);
1008 close(tap_fd);
1009 tap_fd = -1;
1010 *tap_unit = -1;
1011 }
1012 return tap_fd;
1013}
1014
1015static int
1016unix_connect(const char *path)
1017{
1018 struct sockaddr_un sunx;
1019 int len;
1020 int net_fd;
1021 int sndbuf = 262144;
1022 struct stat st;
1023
1024 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1025 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1026 ++len; /* include nul */
1027 sunx.sun_family = AF_UNIX;
1028 sunx.sun_len = len;
1029
1030 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1031 if (net_fd < 0)
1032 return(-1);
1033 if (connect(net_fd, (void *)&sunx, len) < 0) {
1034 close(net_fd);
1035 return(-1);
1036 }
1037 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1038 if (fstat(net_fd, &st) == 0)
1039 printf("Network socket buffer: %d bytes\n", st.st_blksize);
1040 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1041 return(net_fd);
1042}
1043
1044#undef TAPDEV_MAJOR
1045#undef TAPDEV_MINOR
1046#undef TAPDEV_OFLAGS
1047
1048/*
1049 * Following syntax is supported,
1050 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1051 *
1052 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1053 * tap(4)'s netmask len is z
1054 *
1055 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1056 * pseudo netif's address is y.y.y.y
1057 *
1058 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1059 * pseudo netif's address is y.y.y.y
1060 * tap(4) and pseudo netif's netmask len are z
1061 *
1062 * 5) bridgeX tap(4) will be added to bridgeX
1063 *
1064 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1065 * pseudo netif's address is y.y.y.y
1066 *
1067 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1068 * pseudo netif's address is y.y.y.y
1069 * pseudo netif's netmask len is z
1070 */
1071static
1072int
1073netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1074{
1075 in_addr_t tap_addr, netmask, netif_addr;
1076 int next_netif_addr;
1077 char *tok, *masklen_str, *ifbridge;
1078
1079 *addr = 0;
1080 *mask = 0;
1081
1082 tok = strtok(NULL, ":/");
1083 if (tok == NULL) {
1084 /*
1085 * Nothing special, simply use tap(4) as backend
1086 */
1087 return 0;
1088 }
1089
1090 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1091 /*
1092 * tap(4)'s address is supplied
1093 */
1094 ifbridge = NULL;
1095
1096 /*
1097 * If there is next token, then it may be pseudo
1098 * netif's address or netmask len for tap(4)
1099 */
1100 next_netif_addr = 0;
1101 } else {
1102 /*
1103 * Not tap(4)'s address, assume it as a bridge(4)
1104 * iface name
1105 */
1106 tap_addr = 0;
1107 ifbridge = tok;
1108
1109 /*
1110 * If there is next token, then it must be pseudo
1111 * netif's address
1112 */
1113 next_netif_addr = 1;
1114 }
1115
1116 netmask = netif_addr = 0;
1117
1118 tok = strtok(NULL, ":/");
1119 if (tok == NULL)
1120 goto back;
1121
1122 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1123 if (next_netif_addr) {
1124 warnx("Invalid pseudo netif address: %s", tok);
1125 return -1;
1126 }
1127 netif_addr = 0;
1128
1129 /*
1130 * Current token is not address, then it must be netmask len
1131 */
1132 masklen_str = tok;
1133 } else {
1134 /*
1135 * Current token is pseudo netif address, if there is next token
1136 * it must be netmask len
1137 */
1138 masklen_str = strtok(NULL, "/");
1139 }
1140
1141 /* Calculate netmask */
1142 if (masklen_str != NULL) {
1143 u_long masklen;
1144
1145 masklen = strtoul(masklen_str, NULL, 10);
1146 if (masklen < 32 && masklen > 0) {
1147 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1148 & 0xffffffff);
1149 } else {
1150 warnx("Invalid netmask len: %lu", masklen);
1151 return -1;
1152 }
1153 }
1154
1155 /* Make sure there is no more token left */
1156 if (strtok(NULL, ":/") != NULL) {
1157 warnx("Invalid argument to '-I'");
1158 return -1;
1159 }
1160
1161back:
1162 if (tap_unit < 0) {
1163 /* Do nothing */
1164 } else if (ifbridge == NULL) {
1165 /* Set tap(4) address/netmask */
1166 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1167 return -1;
1168 } else {
1169 /* Tie tap(4) to bridge(4) */
1170 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1171 return -1;
1172 }
1173
1174 *addr = netif_addr;
1175 *mask = netmask;
1176 return 0;
1177}
1178
1179/*
1180 * NetifInfo[] will be filled for pseudo netif initialization.
1181 * NetifNum will be bumped to reflect the number of valid entries
1182 * in NetifInfo[].
1183 */
1184static
1185void
1186init_netif(char *netifExp[], int netifExpNum)
1187{
1188 int i, s;
1189
1190 if (netifExpNum == 0)
1191 return;
1192
1193 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1194 if (s < 0)
1195 return;
1196
1197 for (i = 0; i < netifExpNum; ++i) {
1198 struct vknetif_info *info;
1199 in_addr_t netif_addr, netif_mask;
1200 int tap_fd, tap_unit;
1201 char *netif;
1202
1203 netif = strtok(netifExp[i], ":");
1204 if (netif == NULL) {
1205 warnx("Invalid argument to '-I'");
1206 continue;
1207 }
1208
1209 /*
1210 * Open tap(4) device file and bring up the
1211 * corresponding interface
1212 */
1213 tap_fd = netif_open_tap(netif, &tap_unit, s);
1214 if (tap_fd < 0)
1215 continue;
1216
1217 /*
1218 * Initialize tap(4) and get address/netmask
1219 * for pseudo netif
1220 *
1221 * NB: Rest part of netifExp[i] is passed
1222 * to netif_init_tap() implicitly.
1223 */
1224 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1225 /*
1226 * NB: Closing tap(4) device file will bring
1227 * down the corresponding interface
1228 */
1229 close(tap_fd);
1230 continue;
1231 }
1232
1233 info = &NetifInfo[NetifNum];
1234 info->tap_fd = tap_fd;
1235 info->tap_unit = tap_unit;
1236 info->netif_addr = netif_addr;
1237 info->netif_mask = netif_mask;
1238
1239 NetifNum++;
1240 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1241 break;
1242 }
1243 close(s);
1244}
1245
bc3cc25e
MD
1246/*
1247 * Create the pid file and leave it open and locked while the vkernel is
1248 * running. This allows a script to use /usr/bin/lockf to probe whether
1249 * a vkernel is still running (so as not to accidently kill an unrelated
1250 * process from a stale pid file).
1251 */
da673940
JG
1252static
1253void
bc3cc25e 1254writepid(void)
da673940 1255{
bc3cc25e
MD
1256 char buf[32];
1257 int fd;
da673940
JG
1258
1259 if (pid_file != NULL) {
bc3cc25e
MD
1260 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1261 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1262 if (fd < 0) {
1263 if (errno == EWOULDBLOCK) {
1264 perror("Failed to lock pidfile, "
1265 "vkernel already running");
1266 } else {
1267 perror("Failed to create pidfile");
1268 }
1269 exit(EX_SOFTWARE);
da673940 1270 }
bc3cc25e
MD
1271 ftruncate(fd, 0);
1272 write(fd, buf, strlen(buf));
1273 /* leave the file open to maintain the lock */
da673940
JG
1274 }
1275}
1276
1277static
1278void
1279cleanpid( void )
1280{
1281 if (pid_file != NULL) {
bc3cc25e 1282 if (unlink(pid_file) < 0)
da673940
JG
1283 perror("Warning: couldn't remove pidfile");
1284 }
1285}
1286
1287static
1288void
278c4e67 1289usage_err(const char *ctl, ...)
da673940
JG
1290{
1291 va_list va;
1292
1293 va_start(va, ctl);
1294 vfprintf(stderr, ctl, va);
1295 va_end(va);
1296 fprintf(stderr, "\n");
278c4e67
SG
1297 exit(EX_USAGE);
1298}
1299
1300static
1301void
1302usage_help(_Bool help)
1303{
1304 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1305 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
9bea6114
MC
1306 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1307 "\t[-p file] [-r file]\n", save_av[0]);
278c4e67
SG
1308
1309 if (help)
1310 fprintf(stderr, "\nArguments:\n"
1311 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1312 "\t-e\tSpecify an environment to be used by the kernel.\n"
1313 "\t-h\tThis list of options.\n"
1314 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1315 "\t-I\tCreate a virtual network device.\n"
1316 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1317 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
9bea6114
MC
1318 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1319 "\t \t- numcpus - number of cpus\n"
1320 "\t \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1321 "\t \t the logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n"
1322 "\t \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1323 "\t \t the core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n"
278c4e67
SG
1324 "\t-p\tSpecify a file in which to store the process ID.\n"
1325 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1326 "\t-s\tBoot into single-user mode.\n"
1327 "\t-U\tEnable writing to kernel memory and module loading.\n"
1328 "\t-v\tTurn on verbose booting.\n");
1329
1330 exit(EX_USAGE);
da673940
JG
1331}
1332
1333void
1334cpu_reset(void)
1335{
1336 kprintf("cpu reset, rebooting vkernel\n");
1337 closefrom(3);
1338 cleanpid();
1339 execv(save_av[0], save_av);
1340}
1341
1342void
1343cpu_halt(void)
1344{
1345 kprintf("cpu halt, exiting vkernel\n");
1346 cleanpid();
278c4e67 1347 exit(EX_OK);
da673940
JG
1348}
1349
1350void
1351setrealcpu(void)
1352{
1353 switch(lwp_cpu_lock) {
1354 case LCL_PER_CPU:
1355 if (bootverbose)
1356 kprintf("Locking CPU%d to real cpu %d\n",
1357 mycpuid, next_cpu);
1358 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1359 next_cpu++;
1360 if (next_cpu >= real_ncpus)
1361 next_cpu = 0;
1362 break;
1363 case LCL_SINGLE_CPU:
1364 if (bootverbose)
1365 kprintf("Locking CPU%d to real cpu %d\n",
1366 mycpuid, next_cpu);
1367 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1368 break;
1369 default:
1370 /* do not map virtual cpus to real cpus */
1371 break;
1372 }
1373}
bf7c5253
MD
1374
1375/*
1376 * Allocate and free memory for module loading. The loaded module
1377 * has to be placed somewhere near the current kernel binary load
1378 * point or the relocations will not work.
1379 *
1380 * I'm not sure why this isn't working.
1381 */
1382int
1383vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes)
1384{
1385 kprintf("module loading for vkernel64's not currently supported\n");
1386 *basep = 0;
1387 return ENOMEM;
1388#if 0
1389#if 1
1390 size_t xtra;
1391 xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK;
1392 *basep = (vm_offset_t)sbrk(xtra + bytes) + xtra;
1393 bzero((void *)*basep, bytes);
1394#else
1395 *basep = (vm_offset_t)mmap((void *)0x000000000, bytes,
1396 PROT_READ|PROT_WRITE|PROT_EXEC,
1397 MAP_ANON|MAP_SHARED, -1, 0);
1398 if ((void *)*basep == MAP_FAILED)
1399 return ENOMEM;
1400#endif
1401 kprintf("basep %p %p %zd\n",
1402 (void *)vkernel_module_memory_alloc, (void *)*basep, bytes);
1403 return 0;
1404#endif
1405}
1406
1407void
1408vkernel_module_memory_free(vm_offset_t base, size_t bytes)
1409{
1410#if 0
1411#if 0
1412 munmap((void *)base, bytes);
1413#endif
1414#endif
1415}