2 * Copyright (c) 2006 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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
34 * $DragonFly: src/sys/platform/vkernel/platform/init.c,v 1.44 2007/07/10 18:35:38 josepht Exp $
37 #include <sys/types.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
43 #include <sys/random.h>
44 #include <sys/vkernel.h>
46 #include <sys/reboot.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 <vm/vm_page.h>
55 #include <machine/cpu.h>
56 #include <machine/globaldata.h>
57 #include <machine/tls.h>
58 #include <machine/md_var.h>
59 #include <machine/vmparam.h>
62 #include <net/if_arp.h>
63 #include <net/ethernet.h>
64 #include <net/bridge/if_bridgevar.h>
65 #include <netinet/in.h>
66 #include <arpa/inet.h>
78 vm_paddr_t phys_avail[16];
80 vm_paddr_t Maxmem_bytes;
82 struct vkdisk_info DiskInfo[VKDISK_MAX];
84 struct vknetif_info NetifInfo[VKNETIF_MAX];
90 vm_offset_t virtual_start;
91 vm_offset_t virtual_end;
92 vm_offset_t kernel_vm_end;
93 vm_offset_t crashdumpmap;
94 vm_offset_t clean_sva;
95 vm_offset_t clean_eva;
96 struct msgbuf *msgbufp;
99 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
100 u_int cpu_feature; /* XXX */
101 u_int tsc_present; /* XXX */
102 int optcpus; /* number of cpus - see mp_start() */
103 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
104 int real_ncpus; /* number of real CPUs */
105 int next_cpu; /* next real CPU to lock a virtual CPU to */
107 struct privatespace *CPU_prvspace;
109 static struct trapframe proc0_tf;
110 static void *proc0paddr;
112 static void init_sys_memory(char *imageFile);
113 static void init_kern_memory(void);
114 static void init_globaldata(void);
115 static void init_vkernel(void);
116 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
117 static void init_netif(char *netifExp[], int netifFileNum);
118 static void writepid( void );
119 static void cleanpid( void );
120 static void usage(const char *ctl, ...);
123 static char **save_av;
126 * Kernel startup for virtual kernels - standard main()
129 main(int ac, char **av)
131 char *memImageFile = NULL;
132 char *netifFile[VKNETIF_MAX];
133 char *diskFile[VKDISK_MAX];
134 char *cdFile[VKDISK_MAX];
137 size_t real_ncpus_size;
138 int netifFileNum = 0;
151 kernel_mem_readonly = 1;
155 lwp_cpu_lock = LCL_NONE;
157 real_ncpus_size = sizeof(real_ncpus);
158 sysctlbyname("hw.ncpu", &real_ncpus, &real_ncpus_size, NULL, 0);
160 while ((c = getopt(ac, av, "c:svl:m:n:r:e:i:p:I:U")) != -1) {
164 * name=value:name=value:name=value...
167 kern_envp = malloc(n + 2);
168 for (i = 0; i < n; ++i) {
169 if (optarg[i] == ':')
172 kern_envp[i] = optarg[i];
178 boothowto |= RB_SINGLE;
184 memImageFile = optarg;
187 if (netifFileNum < VKNETIF_MAX)
188 netifFile[netifFileNum++] = strdup(optarg);
191 if (diskFileNum + cdFileNum < VKDISK_MAX)
192 diskFile[diskFileNum++] = strdup(optarg);
195 if (diskFileNum + cdFileNum < VKDISK_MAX)
196 cdFile[cdFileNum++] = strdup(optarg);
199 Maxmem_bytes = strtoull(optarg, &suffix, 0);
216 usage("Bad maxmem option");
224 if (strncmp("map", optarg, 3) == 0) {
225 lwp_cpu_lock = LCL_PER_CPU;
226 if (*(optarg + 3) == ',')
227 next_cpu = strtol(optarg+4, &endp, 0);
229 usage("Bad target CPU number at '%s'", endp);
230 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
231 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
232 } else if (strncmp("any", optarg, 3) == 0) {
233 lwp_cpu_lock = LCL_NONE;
235 lwp_cpu_lock = LCL_SINGLE_CPU;
236 next_cpu = strtol(optarg, &endp, 0);
238 usage("Bad target CPU number at '%s'", endp);
239 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
240 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
245 * This value is set up by mp_start(), don't just
249 optcpus = strtol(optarg, NULL, 0);
250 if (optcpus < 1 || optcpus > MAXCPU)
251 usage("Bad ncpus, valid range is 1-%d", MAXCPU);
253 if (strtol(optarg, NULL, 0) != 1) {
254 usage("You built a UP vkernel, only 1 cpu!");
263 kernel_mem_readonly = 0;
270 init_sys_memory(memImageFile);
276 init_disk(diskFile, diskFileNum, VKD_DISK);
277 init_disk(cdFile, cdFileNum, VKD_CD);
278 init_netif(netifFile, netifFileNum);
286 * Initialize system memory. This is the virtual kernel's 'RAM'.
290 init_sys_memory(char *imageFile)
297 * Figure out the system memory image size. If an image file was
298 * specified and -m was not specified, use the image file's size.
301 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
302 Maxmem_bytes = (vm_paddr_t)st.st_size;
303 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
305 err(1, "Cannot create new memory file %s unless "
306 "system memory size is specified with -m",
312 * Maxmem must be known at this time
314 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
315 err(1, "Bad maxmem specification: 32MB minimum, "
316 "multiples of %dMB only",
317 SEG_SIZE / 1024 / 1024);
322 * Generate an image file name if necessary, then open/create the
323 * file exclusively locked. Do not allow multiple virtual kernels
324 * to use the same image file.
326 if (imageFile == NULL) {
327 for (i = 0; i < 1000000; ++i) {
328 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
330 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
331 if (fd < 0 && errno == EWOULDBLOCK) {
338 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
340 printf("Using memory file: %s\n", imageFile);
341 if (fd < 0 || fstat(fd, &st) < 0) {
342 err(1, "Unable to open/create %s", imageFile);
347 * Truncate or extend the file as necessary.
349 if (st.st_size > Maxmem_bytes) {
350 ftruncate(fd, Maxmem_bytes);
351 } else if (st.st_size < Maxmem_bytes) {
353 off_t off = st.st_size & ~SEG_MASK;
355 kprintf("%s: Reserving blocks for memory image\n", imageFile);
356 zmem = malloc(SEG_SIZE);
357 bzero(zmem, SEG_SIZE);
358 lseek(fd, off, SEEK_SET);
359 while (off < Maxmem_bytes) {
360 if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
361 err(1, "Unable to reserve blocks for memory image");
367 err(1, "Unable to reserve blocks for memory image");
371 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
375 * Initialize kernel memory. This reserves kernel virtual memory by using
381 init_kern_memory(void)
387 char *topofstack = &dummy;
392 * Memory map our kernel virtual memory space. Note that the
393 * kernel image itself is not made part of this memory for the
396 * The memory map must be segment-aligned so we can properly
399 * If the system kernel has a different MAXDSIZ, it might not
400 * be possible to map kernel memory in its prefered location.
401 * Try a number of different locations.
403 try = (void *)0x40000000;
405 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
406 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
407 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
411 if (base != MAP_FAILED)
412 munmap(base, KERNEL_KVA_SIZE);
413 try = (char *)try + 0x10000000;
416 err(1, "Unable to mmap() kernel virtual memory!");
419 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
420 KvaStart = (vm_offset_t)base;
421 KvaSize = KERNEL_KVA_SIZE;
422 KvaEnd = KvaStart + KvaSize;
423 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
426 * Create a top-level page table self-mapping itself.
428 * Initialize the page directory at physical page index 0 to point
429 * to an array of page table pages starting at physical page index 1
431 lseek(MemImageFd, 0L, 0);
432 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
433 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
434 write(MemImageFd, &pte, sizeof(pte));
438 * Initialize the PTEs in the page table pages required to map the
439 * page table itself. This includes mapping the page directory page
440 * at the base so we go one more loop then normal.
442 lseek(MemImageFd, PAGE_SIZE, 0);
443 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
444 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
445 write(MemImageFd, &pte, sizeof(pte));
449 * Initialize remaining PTEs to 0. We may be reusing a memory image
450 * file. This is approximately a megabyte.
452 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
453 zero = malloc(PAGE_SIZE);
454 bzero(zero, PAGE_SIZE);
456 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
457 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
462 * Enable the page table and calculate pointers to our self-map
463 * for easy kernel page table manipulation.
465 * KernelPTA must be offset so we can do direct VA translations
467 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
468 0 | VPTE_R | VPTE_W | VPTE_V);
469 KernelPTD = (vpte_t *)base; /* pg directory */
470 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
471 KernelPTA -= KvaStart >> PAGE_SHIFT;
474 * phys_avail[] represents unallocated physical memory. MI code
475 * will use phys_avail[] to create the vm_page array.
477 phys_avail[0] = PAGE_SIZE +
478 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
479 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
480 phys_avail[1] = Maxmem_bytes;
483 * (virtual_start, virtual_end) represent unallocated kernel virtual
484 * memory. MI code will create kernel_map using these parameters.
486 virtual_start = KvaStart + PAGE_SIZE +
487 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
488 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
489 virtual_end = KvaStart + KERNEL_KVA_SIZE;
492 * kernel_vm_end could be set to virtual_end but we want some
493 * indication of how much of the kernel_map we've used, so
494 * set it low and let pmap_growkernel increase it even though we
495 * don't need to create any new page table pages.
497 kernel_vm_end = virtual_start;
500 * Allocate space for process 0's UAREA.
502 proc0paddr = (void *)virtual_start;
503 for (i = 0; i < UPAGES; ++i) {
504 pmap_kenter_quick(virtual_start, phys_avail[0]);
505 virtual_start += PAGE_SIZE;
506 phys_avail[0] += PAGE_SIZE;
512 crashdumpmap = virtual_start;
513 virtual_start += MAXDUMPPGS * PAGE_SIZE;
516 * msgbufp maps the system message buffer
518 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
519 msgbufp = (void *)virtual_start;
520 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
521 pmap_kenter_quick(virtual_start, phys_avail[0]);
522 virtual_start += PAGE_SIZE;
523 phys_avail[0] += PAGE_SIZE;
525 msgbufinit(msgbufp, MSGBUF_SIZE);
528 * used by kern_memio for /dev/mem access
530 ptvmmap = (caddr_t)virtual_start;
531 virtual_start += PAGE_SIZE;
534 * Bootstrap the kernel_pmap
540 * Map the per-cpu globaldata for cpu #0. Allocate the space using
541 * virtual_start and phys_avail[0]
545 init_globaldata(void)
552 * Reserve enough KVA to cover possible cpus. This is a considerable
553 * amount of KVA since the privatespace structure includes two
554 * whole page table mappings.
556 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
557 CPU_prvspace = (void *)virtual_start;
558 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
561 * Allocate enough physical memory to cover the mdglobaldata
562 * portion of the space and the idle stack and map the pages
563 * into KVA. For cpu #0 only.
565 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
567 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
568 pmap_kenter_quick(va, pa);
569 phys_avail[0] += PAGE_SIZE;
571 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
573 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
574 pmap_kenter_quick(va, pa);
575 phys_avail[0] += PAGE_SIZE;
579 * Setup the %gs for cpu #0. The mycpu macro works after this
582 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
586 * Initialize very low level systems including thread0, proc0, etc.
592 struct mdglobaldata *gd;
594 gd = &CPU_prvspace[0].mdglobaldata;
595 bzero(gd, sizeof(*gd));
597 gd->mi.gd_curthread = &thread0;
598 thread0.td_gd = &gd->mi;
600 ncpus2 = 1; /* rounded down power of 2 */
601 ncpus_fit = 1; /* rounded up power of 2 */
602 /* ncpus2_mask and ncpus_fit_mask are 0 */
604 gd->mi.gd_prvspace = &CPU_prvspace[0];
605 mi_gdinit(&gd->mi, 0);
607 mi_proc0init(&gd->mi, proc0paddr);
608 lwp0.lwp_md.md_regs = &proc0_tf;
613 #if 0 /* #ifdef DDB */
615 if (boothowto & RB_KDB)
616 Debugger("Boot flags requested debugger");
619 initializecpu(); /* Initialize CPU registers */
621 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
625 * Map the message buffer
627 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
628 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
629 msgbufinit(msgbufp, MSGBUF_SIZE);
632 thread0.td_pcb_cr3 ... MMU
633 lwp0.lwp_md.md_regs = &proc0_tf;
638 * Filesystem image paths for the virtual kernel are optional.
639 * If specified they each should point to a disk image,
640 * the first of which will become the root disk.
642 * The virtual kernel caches data from our 'disk' just like a normal kernel,
643 * so we do not really want the real kernel to cache the data too. Use
644 * O_DIRECT to remove the duplication.
648 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
652 if (diskFileNum == 0)
655 for(i=0; i < diskFileNum; i++){
660 warnx("Invalid argument to '-r'");
664 if (DiskNum < VKDISK_MAX) {
666 struct vkdisk_info* info = NULL;
670 if (type == VKD_DISK)
671 fd = open(fname, O_RDWR|O_DIRECT|O_EXLOCK|O_NONBLOCK, 0644);
673 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
674 if (fd < 0 || fstat(fd, &st) < 0) {
676 fprintf(stderr, "You may already have a vkernel using this disk image!\n");
677 err(1, "Unable to open/create %s", fname);
680 /* get rid of O_NONBLOCK, keep O_DIRECT */
681 if (type == VKD_DISK)
682 fcntl(fd, F_SETFL, O_DIRECT);
684 info = &DiskInfo[DiskNum];
690 memcpy(info->fname, fname, l);
694 rootdevnames[0] = "cd9660:vcd0a";
695 else if (type == VKD_DISK)
696 rootdevnames[0] = "ufs:vkd0s0a";
701 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
709 netif_set_tapflags(int tap_unit, int f, int s)
714 bzero(&ifr, sizeof(ifr));
716 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
717 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
718 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
725 * If the flags are already set/cleared, then we return
726 * immediately to avoid extra syscalls
728 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
732 if ((flags & f) == 0)
743 * Fix up ifreq.ifr_name, since it may be trashed
744 * in previous ioctl(SIOCGIFFLAGS)
746 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
748 ifr.ifr_flags = flags & 0xffff;
749 ifr.ifr_flagshigh = flags >> 16;
750 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
751 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
759 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
761 struct ifaliasreq ifra;
762 struct sockaddr_in *in;
764 bzero(&ifra, sizeof(ifra));
765 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
768 in = (struct sockaddr_in *)&ifra.ifra_addr;
769 in->sin_family = AF_INET;
770 in->sin_len = sizeof(*in);
771 in->sin_addr.s_addr = addr;
775 in = (struct sockaddr_in *)&ifra.ifra_mask;
776 in->sin_len = sizeof(*in);
777 in->sin_addr.s_addr = mask;
780 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
781 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
789 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
794 bzero(&ifbr, sizeof(ifbr));
795 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
798 bzero(&ifd, sizeof(ifd));
799 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
800 ifd.ifd_cmd = BRDGADD;
801 ifd.ifd_len = sizeof(ifbr);
802 ifd.ifd_data = &ifbr;
804 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
806 * 'errno == EEXIST' means that the tap(4) is already
807 * a member of the bridge(4)
809 if (errno != EEXIST) {
810 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
817 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
819 /* XXX major()/minor() can't be used in vkernel */
820 #define TAPDEV_MAJOR(x) ((int)(((u_int)(x) >> 8) & 0xff))
821 #define TAPDEV_MINOR(x) ((int)((x) & 0xffff00ff))
823 #ifndef TAP_CDEV_MAJOR
824 #define TAP_CDEV_MAJOR 149
828 * Locate the first unused tap(4) device file if auto mode is requested,
829 * or open the user supplied device file, and bring up the corresponding
832 * NOTE: Only tap(4) device file is supported currently
836 netif_open_tap(const char *netif, int *tap_unit, int s)
838 char tap_dev[MAXPATHLEN];
844 if (strcmp(netif, "auto") == 0) {
848 * Find first unused tap(4) device file
851 snprintf(tap_dev, sizeof(tap_dev), "/dev/tap%d", i);
852 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
853 if (tap_fd >= 0 || errno == ENOENT)
857 warnx("Unable to find a free tap(4)");
862 * User supplied tap(4) device file
864 if (netif[0] == '/') /* Absolute path */
865 strlcpy(tap_dev, netif, sizeof(tap_dev));
867 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
869 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
871 warn("Unable to open %s", tap_dev);
877 * Check whether the device file is a tap(4)
880 if (fstat(tap_fd, &st) == 0 && S_ISCHR(st.st_mode) &&
881 TAPDEV_MAJOR(st.st_rdev) == TAP_CDEV_MAJOR) {
882 *tap_unit = TAPDEV_MINOR(st.st_rdev);
885 * Bring up the corresponding tap(4) interface
887 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
890 warnx("%s is not a tap(4) device", tap_dev);
906 * Following syntax is supported,
907 * 1) x.x.x.x tap(4)'s address is x.x.x.x
909 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
910 * tap(4)'s netmask len is z
912 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
913 * pseudo netif's address is y.y.y.y
915 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
916 * pseudo netif's address is y.y.y.y
917 * tap(4) and pseudo netif's netmask len are z
919 * 5) bridgeX tap(4) will be added to bridgeX
921 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
922 * pseudo netif's address is y.y.y.y
924 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
925 * pseudo netif's address is y.y.y.y
926 * pseudo netif's netmask len is z
930 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
932 in_addr_t tap_addr, netmask, netif_addr;
934 char *tok, *masklen_str, *ifbridge;
939 tok = strtok(NULL, ":/");
942 * Nothing special, simply use tap(4) as backend
947 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
949 * tap(4)'s address is supplied
954 * If there is next token, then it may be pseudo
955 * netif's address or netmask len for tap(4)
960 * Not tap(4)'s address, assume it as a bridge(4)
967 * If there is next token, then it must be pseudo
973 netmask = netif_addr = 0;
975 tok = strtok(NULL, ":/");
979 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
980 if (next_netif_addr) {
981 warnx("Invalid pseudo netif address: %s", tok);
987 * Current token is not address, then it must be netmask len
992 * Current token is pseudo netif address, if there is next token
993 * it must be netmask len
995 masklen_str = strtok(NULL, "/");
998 /* Calculate netmask */
999 if (masklen_str != NULL) {
1002 masklen = strtoul(masklen_str, NULL, 10);
1003 if (masklen < 32 && masklen > 0) {
1004 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1007 warnx("Invalid netmask len: %lu", masklen);
1012 /* Make sure there is no more token left */
1013 if (strtok(NULL, ":/") != NULL) {
1014 warnx("Invalid argument to '-I'");
1019 if (ifbridge == NULL) {
1020 /* Set tap(4) address/netmask */
1021 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1024 /* Tie tap(4) to bridge(4) */
1025 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1035 * NetifInfo[] will be filled for pseudo netif initialization.
1036 * NetifNum will be bumped to reflect the number of valid entries
1041 init_netif(char *netifExp[], int netifExpNum)
1045 if (netifExpNum == 0)
1048 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1052 for (i = 0; i < netifExpNum; ++i) {
1053 struct vknetif_info *info;
1054 in_addr_t netif_addr, netif_mask;
1055 int tap_fd, tap_unit;
1058 netif = strtok(netifExp[i], ":");
1059 if (netif == NULL) {
1060 warnx("Invalid argument to '-I'");
1065 * Open tap(4) device file and bring up the
1066 * corresponding interface
1068 tap_fd = netif_open_tap(netif, &tap_unit, s);
1073 * Initialize tap(4) and get address/netmask
1076 * NB: Rest part of netifExp[i] is passed
1077 * to netif_init_tap() implicitly.
1079 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1081 * NB: Closing tap(4) device file will bring
1082 * down the corresponding interface
1088 info = &NetifInfo[NetifNum];
1089 info->tap_fd = tap_fd;
1090 info->tap_unit = tap_unit;
1091 info->netif_addr = netif_addr;
1092 info->netif_mask = netif_mask;
1095 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1108 if (pid_file != NULL) {
1110 fp = fopen(pid_file, "w");
1113 fprintf(fp, "%ld\n", (long)self);
1117 perror("Warning: couldn't open pidfile");
1126 if (pid_file != NULL) {
1127 if ( unlink(pid_file) != 0 )
1128 perror("Warning: couldn't remove pidfile");
1134 usage(const char *ctl, ...)
1139 vfprintf(stderr, ctl, va);
1141 fprintf(stderr, "\n");
1148 kprintf("cpu reset, rebooting vkernel\n");
1151 execv(save_av[0], save_av);
1157 kprintf("cpu halt, exiting vkernel\n");
1165 switch(lwp_cpu_lock) {
1168 kprintf("Locking CPU%d to real cpu %d\n",
1170 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1172 if (next_cpu >= real_ncpus)
1175 case LCL_SINGLE_CPU:
1177 kprintf("Locking CPU%d to real cpu %d\n",
1179 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1182 /* do not map virtual cpus to real cpus */