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 1.47 2007/08/20 05:27:46 dillon 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;
141 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
145 int real_vkernel_enable;
146 size_t real_vkernel_enable_size;
154 kernel_mem_readonly = 1;
158 lwp_cpu_lock = LCL_NONE;
160 real_vkernel_enable_size = sizeof(real_vkernel_enable);
161 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &real_vkernel_enable_size, NULL, 0);
163 if (real_vkernel_enable == 0)
164 errx(1,"vm.vkernel_enable is %d, must be set to 1 to execute a vkernel!", real_vkernel_enable);
166 real_ncpus_size = sizeof(real_ncpus);
167 sysctlbyname("hw.ncpu", &real_ncpus, &real_ncpus_size, NULL, 0);
169 while ((c = getopt(ac, av, "c:svl:m:n:r:e:i:p:I:U")) != -1) {
173 * name=value:name=value:name=value...
176 kern_envp = malloc(n + 2);
177 for (i = 0; i < n; ++i) {
178 if (optarg[i] == ':')
181 kern_envp[i] = optarg[i];
187 boothowto |= RB_SINGLE;
193 memImageFile = optarg;
196 if (netifFileNum < VKNETIF_MAX)
197 netifFile[netifFileNum++] = strdup(optarg);
202 if (diskFileNum + cdFileNum < VKDISK_MAX)
203 diskFile[diskFileNum++] = strdup(optarg);
208 if (diskFileNum + cdFileNum < VKDISK_MAX)
209 cdFile[cdFileNum++] = strdup(optarg);
212 Maxmem_bytes = strtoull(optarg, &suffix, 0);
229 usage("Bad maxmem option");
237 if (strncmp("map", optarg, 3) == 0) {
238 lwp_cpu_lock = LCL_PER_CPU;
239 if (optarg[3] == ',') {
240 next_cpu = strtol(optarg+4, &endp, 0);
242 usage("Bad target CPU number at '%s'", endp);
246 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
247 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
248 } else if (strncmp("any", optarg, 3) == 0) {
249 lwp_cpu_lock = LCL_NONE;
251 lwp_cpu_lock = LCL_SINGLE_CPU;
252 next_cpu = strtol(optarg, &endp, 0);
254 usage("Bad target CPU number at '%s'", endp);
255 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
256 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
261 * This value is set up by mp_start(), don't just
265 optcpus = strtol(optarg, NULL, 0);
266 if (optcpus < 1 || optcpus > MAXCPU)
267 usage("Bad ncpus, valid range is 1-%d", MAXCPU);
269 if (strtol(optarg, NULL, 0) != 1) {
270 usage("You built a UP vkernel, only 1 cpu!");
279 kernel_mem_readonly = 0;
286 init_sys_memory(memImageFile);
294 * We boot from the first installed disk.
296 if (bootOnDisk == 1) {
297 init_disk(diskFile, diskFileNum, VKD_DISK);
298 init_disk(cdFile, cdFileNum, VKD_CD);
300 init_disk(cdFile, cdFileNum, VKD_CD);
301 init_disk(diskFile, diskFileNum, VKD_DISK);
303 init_netif(netifFile, netifFileNum);
311 * Initialize system memory. This is the virtual kernel's 'RAM'.
315 init_sys_memory(char *imageFile)
322 * Figure out the system memory image size. If an image file was
323 * specified and -m was not specified, use the image file's size.
326 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
327 Maxmem_bytes = (vm_paddr_t)st.st_size;
328 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
330 err(1, "Cannot create new memory file %s unless "
331 "system memory size is specified with -m",
337 * Maxmem must be known at this time
339 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
340 err(1, "Bad maxmem specification: 32MB minimum, "
341 "multiples of %dMB only",
342 SEG_SIZE / 1024 / 1024);
347 * Generate an image file name if necessary, then open/create the
348 * file exclusively locked. Do not allow multiple virtual kernels
349 * to use the same image file.
351 if (imageFile == NULL) {
352 for (i = 0; i < 1000000; ++i) {
353 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
355 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
356 if (fd < 0 && errno == EWOULDBLOCK) {
363 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
365 printf("Using memory file: %s\n", imageFile);
366 if (fd < 0 || fstat(fd, &st) < 0) {
367 err(1, "Unable to open/create %s", imageFile);
372 * Truncate or extend the file as necessary.
374 if (st.st_size > Maxmem_bytes) {
375 ftruncate(fd, Maxmem_bytes);
376 } else if (st.st_size < Maxmem_bytes) {
378 off_t off = st.st_size & ~SEG_MASK;
380 kprintf("%s: Reserving blocks for memory image\n", imageFile);
381 zmem = malloc(SEG_SIZE);
382 bzero(zmem, SEG_SIZE);
383 lseek(fd, off, SEEK_SET);
384 while (off < Maxmem_bytes) {
385 if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
386 err(1, "Unable to reserve blocks for memory image");
392 err(1, "Unable to reserve blocks for memory image");
396 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
400 * Initialize kernel memory. This reserves kernel virtual memory by using
406 init_kern_memory(void)
412 char *topofstack = &dummy;
417 * Memory map our kernel virtual memory space. Note that the
418 * kernel image itself is not made part of this memory for the
421 * The memory map must be segment-aligned so we can properly
424 * If the system kernel has a different MAXDSIZ, it might not
425 * be possible to map kernel memory in its prefered location.
426 * Try a number of different locations.
428 try = (void *)0x40000000;
430 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
431 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
432 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
436 if (base != MAP_FAILED)
437 munmap(base, KERNEL_KVA_SIZE);
438 try = (char *)try + 0x10000000;
441 err(1, "Unable to mmap() kernel virtual memory!");
444 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
445 KvaStart = (vm_offset_t)base;
446 KvaSize = KERNEL_KVA_SIZE;
447 KvaEnd = KvaStart + KvaSize;
448 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
451 * Create a top-level page table self-mapping itself.
453 * Initialize the page directory at physical page index 0 to point
454 * to an array of page table pages starting at physical page index 1
456 lseek(MemImageFd, 0L, 0);
457 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
458 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
459 write(MemImageFd, &pte, sizeof(pte));
463 * Initialize the PTEs in the page table pages required to map the
464 * page table itself. This includes mapping the page directory page
465 * at the base so we go one more loop then normal.
467 lseek(MemImageFd, PAGE_SIZE, 0);
468 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
469 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
470 write(MemImageFd, &pte, sizeof(pte));
474 * Initialize remaining PTEs to 0. We may be reusing a memory image
475 * file. This is approximately a megabyte.
477 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
478 zero = malloc(PAGE_SIZE);
479 bzero(zero, PAGE_SIZE);
481 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
482 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
487 * Enable the page table and calculate pointers to our self-map
488 * for easy kernel page table manipulation.
490 * KernelPTA must be offset so we can do direct VA translations
492 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
493 0 | VPTE_R | VPTE_W | VPTE_V);
494 KernelPTD = (vpte_t *)base; /* pg directory */
495 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
496 KernelPTA -= KvaStart >> PAGE_SHIFT;
499 * phys_avail[] represents unallocated physical memory. MI code
500 * will use phys_avail[] to create the vm_page array.
502 phys_avail[0] = PAGE_SIZE +
503 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
504 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
505 phys_avail[1] = Maxmem_bytes;
508 * (virtual_start, virtual_end) represent unallocated kernel virtual
509 * memory. MI code will create kernel_map using these parameters.
511 virtual_start = KvaStart + PAGE_SIZE +
512 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
513 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
514 virtual_end = KvaStart + KERNEL_KVA_SIZE;
517 * kernel_vm_end could be set to virtual_end but we want some
518 * indication of how much of the kernel_map we've used, so
519 * set it low and let pmap_growkernel increase it even though we
520 * don't need to create any new page table pages.
522 kernel_vm_end = virtual_start;
525 * Allocate space for process 0's UAREA.
527 proc0paddr = (void *)virtual_start;
528 for (i = 0; i < UPAGES; ++i) {
529 pmap_kenter_quick(virtual_start, phys_avail[0]);
530 virtual_start += PAGE_SIZE;
531 phys_avail[0] += PAGE_SIZE;
537 crashdumpmap = virtual_start;
538 virtual_start += MAXDUMPPGS * PAGE_SIZE;
541 * msgbufp maps the system message buffer
543 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
544 msgbufp = (void *)virtual_start;
545 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
546 pmap_kenter_quick(virtual_start, phys_avail[0]);
547 virtual_start += PAGE_SIZE;
548 phys_avail[0] += PAGE_SIZE;
550 msgbufinit(msgbufp, MSGBUF_SIZE);
553 * used by kern_memio for /dev/mem access
555 ptvmmap = (caddr_t)virtual_start;
556 virtual_start += PAGE_SIZE;
559 * Bootstrap the kernel_pmap
565 * Map the per-cpu globaldata for cpu #0. Allocate the space using
566 * virtual_start and phys_avail[0]
570 init_globaldata(void)
577 * Reserve enough KVA to cover possible cpus. This is a considerable
578 * amount of KVA since the privatespace structure includes two
579 * whole page table mappings.
581 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
582 CPU_prvspace = (void *)virtual_start;
583 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
586 * Allocate enough physical memory to cover the mdglobaldata
587 * portion of the space and the idle stack and map the pages
588 * into KVA. For cpu #0 only.
590 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
592 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
593 pmap_kenter_quick(va, pa);
594 phys_avail[0] += PAGE_SIZE;
596 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
598 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
599 pmap_kenter_quick(va, pa);
600 phys_avail[0] += PAGE_SIZE;
604 * Setup the %gs for cpu #0. The mycpu macro works after this
607 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
611 * Initialize very low level systems including thread0, proc0, etc.
617 struct mdglobaldata *gd;
619 gd = &CPU_prvspace[0].mdglobaldata;
620 bzero(gd, sizeof(*gd));
622 gd->mi.gd_curthread = &thread0;
623 thread0.td_gd = &gd->mi;
625 ncpus2 = 1; /* rounded down power of 2 */
626 ncpus_fit = 1; /* rounded up power of 2 */
627 /* ncpus2_mask and ncpus_fit_mask are 0 */
629 gd->mi.gd_prvspace = &CPU_prvspace[0];
630 mi_gdinit(&gd->mi, 0);
632 mi_proc0init(&gd->mi, proc0paddr);
633 lwp0.lwp_md.md_regs = &proc0_tf;
638 #if 0 /* #ifdef DDB */
640 if (boothowto & RB_KDB)
641 Debugger("Boot flags requested debugger");
644 initializecpu(); /* Initialize CPU registers */
646 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
650 * Map the message buffer
652 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
653 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
654 msgbufinit(msgbufp, MSGBUF_SIZE);
657 thread0.td_pcb_cr3 ... MMU
658 lwp0.lwp_md.md_regs = &proc0_tf;
663 * Filesystem image paths for the virtual kernel are optional.
664 * If specified they each should point to a disk image,
665 * the first of which will become the root disk.
667 * The virtual kernel caches data from our 'disk' just like a normal kernel,
668 * so we do not really want the real kernel to cache the data too. Use
669 * O_DIRECT to remove the duplication.
673 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
677 if (diskFileNum == 0)
680 for(i=0; i < diskFileNum; i++){
685 warnx("Invalid argument to '-r'");
689 if (DiskNum < VKDISK_MAX) {
691 struct vkdisk_info* info = NULL;
695 if (type == VKD_DISK)
696 fd = open(fname, O_RDWR|O_DIRECT|O_EXLOCK|O_NONBLOCK, 0644);
698 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
699 if (fd < 0 || fstat(fd, &st) < 0) {
701 fprintf(stderr, "You may already have a vkernel using this disk image!\n");
702 err(1, "Unable to open/create %s", fname);
705 /* get rid of O_NONBLOCK, keep O_DIRECT */
706 if (type == VKD_DISK)
707 fcntl(fd, F_SETFL, O_DIRECT);
709 info = &DiskInfo[DiskNum];
715 memcpy(info->fname, fname, l);
719 rootdevnames[0] = "cd9660:vcd0a";
720 else if (type == VKD_DISK)
721 rootdevnames[0] = "ufs:vkd0s0a";
726 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
734 netif_set_tapflags(int tap_unit, int f, int s)
739 bzero(&ifr, sizeof(ifr));
741 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
742 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
743 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
750 * If the flags are already set/cleared, then we return
751 * immediately to avoid extra syscalls
753 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
757 if ((flags & f) == 0)
768 * Fix up ifreq.ifr_name, since it may be trashed
769 * in previous ioctl(SIOCGIFFLAGS)
771 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
773 ifr.ifr_flags = flags & 0xffff;
774 ifr.ifr_flagshigh = flags >> 16;
775 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
776 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
784 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
786 struct ifaliasreq ifra;
787 struct sockaddr_in *in;
789 bzero(&ifra, sizeof(ifra));
790 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
793 in = (struct sockaddr_in *)&ifra.ifra_addr;
794 in->sin_family = AF_INET;
795 in->sin_len = sizeof(*in);
796 in->sin_addr.s_addr = addr;
800 in = (struct sockaddr_in *)&ifra.ifra_mask;
801 in->sin_len = sizeof(*in);
802 in->sin_addr.s_addr = mask;
805 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
806 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
814 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
819 bzero(&ifbr, sizeof(ifbr));
820 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
823 bzero(&ifd, sizeof(ifd));
824 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
825 ifd.ifd_cmd = BRDGADD;
826 ifd.ifd_len = sizeof(ifbr);
827 ifd.ifd_data = &ifbr;
829 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
831 * 'errno == EEXIST' means that the tap(4) is already
832 * a member of the bridge(4)
834 if (errno != EEXIST) {
835 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
842 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
844 /* XXX major()/minor() can't be used in vkernel */
845 #define TAPDEV_MAJOR(x) ((int)(((u_int)(x) >> 8) & 0xff))
846 #define TAPDEV_MINOR(x) ((int)((x) & 0xffff00ff))
848 #ifndef TAP_CDEV_MAJOR
849 #define TAP_CDEV_MAJOR 149
853 * Locate the first unused tap(4) device file if auto mode is requested,
854 * or open the user supplied device file, and bring up the corresponding
857 * NOTE: Only tap(4) device file is supported currently
861 netif_open_tap(const char *netif, int *tap_unit, int s)
863 char tap_dev[MAXPATHLEN];
869 if (strcmp(netif, "auto") == 0) {
873 * Find first unused tap(4) device file
876 snprintf(tap_dev, sizeof(tap_dev), "/dev/tap%d", i);
877 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
878 if (tap_fd >= 0 || errno == ENOENT)
882 warnx("Unable to find a free tap(4)");
887 * User supplied tap(4) device file
889 if (netif[0] == '/') /* Absolute path */
890 strlcpy(tap_dev, netif, sizeof(tap_dev));
892 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
894 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
896 warn("Unable to open %s", tap_dev);
902 * Check whether the device file is a tap(4)
905 if (fstat(tap_fd, &st) == 0 && S_ISCHR(st.st_mode) &&
906 TAPDEV_MAJOR(st.st_rdev) == TAP_CDEV_MAJOR) {
907 *tap_unit = TAPDEV_MINOR(st.st_rdev);
910 * Bring up the corresponding tap(4) interface
912 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
915 warnx("%s is not a tap(4) device", tap_dev);
931 * Following syntax is supported,
932 * 1) x.x.x.x tap(4)'s address is x.x.x.x
934 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
935 * tap(4)'s netmask len is z
937 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
938 * pseudo netif's address is y.y.y.y
940 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
941 * pseudo netif's address is y.y.y.y
942 * tap(4) and pseudo netif's netmask len are z
944 * 5) bridgeX tap(4) will be added to bridgeX
946 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
947 * pseudo netif's address is y.y.y.y
949 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
950 * pseudo netif's address is y.y.y.y
951 * pseudo netif's netmask len is z
955 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
957 in_addr_t tap_addr, netmask, netif_addr;
959 char *tok, *masklen_str, *ifbridge;
964 tok = strtok(NULL, ":/");
967 * Nothing special, simply use tap(4) as backend
972 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
974 * tap(4)'s address is supplied
979 * If there is next token, then it may be pseudo
980 * netif's address or netmask len for tap(4)
985 * Not tap(4)'s address, assume it as a bridge(4)
992 * If there is next token, then it must be pseudo
998 netmask = netif_addr = 0;
1000 tok = strtok(NULL, ":/");
1004 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1005 if (next_netif_addr) {
1006 warnx("Invalid pseudo netif address: %s", tok);
1012 * Current token is not address, then it must be netmask len
1017 * Current token is pseudo netif address, if there is next token
1018 * it must be netmask len
1020 masklen_str = strtok(NULL, "/");
1023 /* Calculate netmask */
1024 if (masklen_str != NULL) {
1027 masklen = strtoul(masklen_str, NULL, 10);
1028 if (masklen < 32 && masklen > 0) {
1029 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1032 warnx("Invalid netmask len: %lu", masklen);
1037 /* Make sure there is no more token left */
1038 if (strtok(NULL, ":/") != NULL) {
1039 warnx("Invalid argument to '-I'");
1044 if (ifbridge == NULL) {
1045 /* Set tap(4) address/netmask */
1046 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1049 /* Tie tap(4) to bridge(4) */
1050 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1060 * NetifInfo[] will be filled for pseudo netif initialization.
1061 * NetifNum will be bumped to reflect the number of valid entries
1066 init_netif(char *netifExp[], int netifExpNum)
1070 if (netifExpNum == 0)
1073 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1077 for (i = 0; i < netifExpNum; ++i) {
1078 struct vknetif_info *info;
1079 in_addr_t netif_addr, netif_mask;
1080 int tap_fd, tap_unit;
1083 netif = strtok(netifExp[i], ":");
1084 if (netif == NULL) {
1085 warnx("Invalid argument to '-I'");
1090 * Open tap(4) device file and bring up the
1091 * corresponding interface
1093 tap_fd = netif_open_tap(netif, &tap_unit, s);
1098 * Initialize tap(4) and get address/netmask
1101 * NB: Rest part of netifExp[i] is passed
1102 * to netif_init_tap() implicitly.
1104 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1106 * NB: Closing tap(4) device file will bring
1107 * down the corresponding interface
1113 info = &NetifInfo[NetifNum];
1114 info->tap_fd = tap_fd;
1115 info->tap_unit = tap_unit;
1116 info->netif_addr = netif_addr;
1117 info->netif_mask = netif_mask;
1120 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1133 if (pid_file != NULL) {
1135 fp = fopen(pid_file, "w");
1138 fprintf(fp, "%ld\n", (long)self);
1142 perror("Warning: couldn't open pidfile");
1151 if (pid_file != NULL) {
1152 if ( unlink(pid_file) != 0 )
1153 perror("Warning: couldn't remove pidfile");
1159 usage(const char *ctl, ...)
1164 vfprintf(stderr, ctl, va);
1166 fprintf(stderr, "\n");
1173 kprintf("cpu reset, rebooting vkernel\n");
1176 execv(save_av[0], save_av);
1182 kprintf("cpu halt, exiting vkernel\n");
1190 switch(lwp_cpu_lock) {
1193 kprintf("Locking CPU%d to real cpu %d\n",
1195 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1197 if (next_cpu >= real_ncpus)
1200 case LCL_SINGLE_CPU:
1202 kprintf("Locking CPU%d to real cpu %d\n",
1204 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1207 /* do not map virtual cpus to real cpus */