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 1.52 2008/05/10 17:24:12 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>
60 #include <cpu/specialreg.h>
63 #include <net/if_arp.h>
64 #include <net/ethernet.h>
65 #include <net/bridge/if_bridgevar.h>
66 #include <netinet/in.h>
67 #include <arpa/inet.h>
79 vm_paddr_t phys_avail[16];
81 vm_paddr_t Maxmem_bytes;
83 struct vkdisk_info DiskInfo[VKDISK_MAX];
85 struct vknetif_info NetifInfo[VKNETIF_MAX];
91 vm_offset_t virtual_start;
92 vm_offset_t virtual_end;
93 vm_offset_t kernel_vm_end;
94 vm_offset_t crashdumpmap;
95 vm_offset_t clean_sva;
96 vm_offset_t clean_eva;
97 struct msgbuf *msgbufp;
100 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
101 u_int cpu_feature; /* XXX */
103 int64_t tsc_frequency;
104 int optcpus; /* number of cpus - see mp_start() */
105 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
106 int real_ncpus; /* number of real CPUs */
107 int next_cpu; /* next real CPU to lock a virtual CPU to */
109 struct privatespace *CPU_prvspace;
111 static struct trapframe proc0_tf;
112 static void *proc0paddr;
114 static void init_sys_memory(char *imageFile);
115 static void init_kern_memory(void);
116 static void init_globaldata(void);
117 static void init_vkernel(void);
118 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
119 static void init_netif(char *netifExp[], int netifFileNum);
120 static void writepid( void );
121 static void cleanpid( void );
122 static void usage(const char *ctl, ...);
125 static char **save_av;
128 * Kernel startup for virtual kernels - standard main()
131 main(int ac, char **av)
133 char *memImageFile = NULL;
134 char *netifFile[VKNETIF_MAX];
135 char *diskFile[VKDISK_MAX];
136 char *cdFile[VKDISK_MAX];
139 int netifFileNum = 0;
142 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
146 int real_vkernel_enable;
156 kernel_mem_readonly = 1;
160 lwp_cpu_lock = LCL_NONE;
162 real_vkernel_enable = 0;
163 vsize = sizeof(real_vkernel_enable);
164 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
166 if (real_vkernel_enable == 0) {
167 errx(1, "vm.vkernel_enable is 0, must be set "
168 "to 1 to execute a vkernel!");
172 vsize = sizeof(real_ncpus);
173 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
175 while ((c = getopt(ac, av, "c:svl:m:n:r:e:i:p:I:U")) != -1) {
179 * name=value:name=value:name=value...
182 kern_envp = malloc(n + 2);
183 for (i = 0; i < n; ++i) {
184 if (optarg[i] == ':')
187 kern_envp[i] = optarg[i];
193 boothowto |= RB_SINGLE;
199 memImageFile = optarg;
202 if (netifFileNum < VKNETIF_MAX)
203 netifFile[netifFileNum++] = strdup(optarg);
208 if (diskFileNum + cdFileNum < VKDISK_MAX)
209 diskFile[diskFileNum++] = strdup(optarg);
214 if (diskFileNum + cdFileNum < VKDISK_MAX)
215 cdFile[cdFileNum++] = strdup(optarg);
218 Maxmem_bytes = strtoull(optarg, &suffix, 0);
235 usage("Bad maxmem option");
243 if (strncmp("map", optarg, 3) == 0) {
244 lwp_cpu_lock = LCL_PER_CPU;
245 if (optarg[3] == ',') {
246 next_cpu = strtol(optarg+4, &endp, 0);
248 usage("Bad target CPU number at '%s'", endp);
252 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
253 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
254 } else if (strncmp("any", optarg, 3) == 0) {
255 lwp_cpu_lock = LCL_NONE;
257 lwp_cpu_lock = LCL_SINGLE_CPU;
258 next_cpu = strtol(optarg, &endp, 0);
260 usage("Bad target CPU number at '%s'", endp);
261 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
262 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
267 * This value is set up by mp_start(), don't just
271 optcpus = strtol(optarg, NULL, 0);
272 if (optcpus < 1 || optcpus > MAXCPU)
273 usage("Bad ncpus, valid range is 1-%d", MAXCPU);
275 if (strtol(optarg, NULL, 0) != 1) {
276 usage("You built a UP vkernel, only 1 cpu!");
285 kernel_mem_readonly = 0;
292 init_sys_memory(memImageFile);
302 vsize = sizeof(tsc_present);
303 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
304 vsize = sizeof(tsc_frequency);
305 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
307 cpu_feature |= CPUID_TSC;
312 vsize = sizeof(supports_sse);
314 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
315 init_fpu(supports_sse);
317 cpu_feature |= CPUID_SSE | CPUID_FXSR;
320 * We boot from the first installed disk.
322 if (bootOnDisk == 1) {
323 init_disk(diskFile, diskFileNum, VKD_DISK);
324 init_disk(cdFile, cdFileNum, VKD_CD);
326 init_disk(cdFile, cdFileNum, VKD_CD);
327 init_disk(diskFile, diskFileNum, VKD_DISK);
329 init_netif(netifFile, netifFileNum);
337 * Initialize system memory. This is the virtual kernel's 'RAM'.
341 init_sys_memory(char *imageFile)
348 * Figure out the system memory image size. If an image file was
349 * specified and -m was not specified, use the image file's size.
352 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
353 Maxmem_bytes = (vm_paddr_t)st.st_size;
354 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
356 err(1, "Cannot create new memory file %s unless "
357 "system memory size is specified with -m",
363 * Maxmem must be known at this time
365 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
366 err(1, "Bad maxmem specification: 32MB minimum, "
367 "multiples of %dMB only",
368 SEG_SIZE / 1024 / 1024);
373 * Generate an image file name if necessary, then open/create the
374 * file exclusively locked. Do not allow multiple virtual kernels
375 * to use the same image file.
377 if (imageFile == NULL) {
378 for (i = 0; i < 1000000; ++i) {
379 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
381 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
382 if (fd < 0 && errno == EWOULDBLOCK) {
389 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
391 printf("Using memory file: %s\n", imageFile);
392 if (fd < 0 || fstat(fd, &st) < 0) {
393 err(1, "Unable to open/create %s", imageFile);
398 * Truncate or extend the file as necessary.
400 if (st.st_size > Maxmem_bytes) {
401 ftruncate(fd, Maxmem_bytes);
402 } else if (st.st_size < Maxmem_bytes) {
404 off_t off = st.st_size & ~SEG_MASK;
406 kprintf("%s: Reserving blocks for memory image\n", imageFile);
407 zmem = malloc(SEG_SIZE);
408 bzero(zmem, SEG_SIZE);
409 lseek(fd, off, SEEK_SET);
410 while (off < Maxmem_bytes) {
411 if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
412 err(1, "Unable to reserve blocks for memory image");
418 err(1, "Unable to reserve blocks for memory image");
422 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
426 * Initialize kernel memory. This reserves kernel virtual memory by using
432 init_kern_memory(void)
438 char *topofstack = &dummy;
443 * Memory map our kernel virtual memory space. Note that the
444 * kernel image itself is not made part of this memory for the
447 * The memory map must be segment-aligned so we can properly
450 * If the system kernel has a different MAXDSIZ, it might not
451 * be possible to map kernel memory in its prefered location.
452 * Try a number of different locations.
454 try = (void *)0x40000000;
456 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
457 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
458 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
462 if (base != MAP_FAILED)
463 munmap(base, KERNEL_KVA_SIZE);
464 try = (char *)try + 0x10000000;
467 err(1, "Unable to mmap() kernel virtual memory!");
470 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
471 KvaStart = (vm_offset_t)base;
472 KvaSize = KERNEL_KVA_SIZE;
473 KvaEnd = KvaStart + KvaSize;
474 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
477 * Create a top-level page table self-mapping itself.
479 * Initialize the page directory at physical page index 0 to point
480 * to an array of page table pages starting at physical page index 1
482 lseek(MemImageFd, 0L, 0);
483 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
484 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
485 write(MemImageFd, &pte, sizeof(pte));
489 * Initialize the PTEs in the page table pages required to map the
490 * page table itself. This includes mapping the page directory page
491 * at the base so we go one more loop then normal.
493 lseek(MemImageFd, PAGE_SIZE, 0);
494 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
495 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
496 write(MemImageFd, &pte, sizeof(pte));
500 * Initialize remaining PTEs to 0. We may be reusing a memory image
501 * file. This is approximately a megabyte.
503 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
504 zero = malloc(PAGE_SIZE);
505 bzero(zero, PAGE_SIZE);
507 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
508 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
513 * Enable the page table and calculate pointers to our self-map
514 * for easy kernel page table manipulation.
516 * KernelPTA must be offset so we can do direct VA translations
518 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
519 0 | VPTE_R | VPTE_W | VPTE_V);
520 KernelPTD = (vpte_t *)base; /* pg directory */
521 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
522 KernelPTA -= KvaStart >> PAGE_SHIFT;
525 * phys_avail[] represents unallocated physical memory. MI code
526 * will use phys_avail[] to create the vm_page array.
528 phys_avail[0] = PAGE_SIZE +
529 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
530 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
531 phys_avail[1] = Maxmem_bytes;
534 * (virtual_start, virtual_end) represent unallocated kernel virtual
535 * memory. MI code will create kernel_map using these parameters.
537 virtual_start = KvaStart + PAGE_SIZE +
538 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
539 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
540 virtual_end = KvaStart + KERNEL_KVA_SIZE;
543 * kernel_vm_end could be set to virtual_end but we want some
544 * indication of how much of the kernel_map we've used, so
545 * set it low and let pmap_growkernel increase it even though we
546 * don't need to create any new page table pages.
548 kernel_vm_end = virtual_start;
551 * Allocate space for process 0's UAREA.
553 proc0paddr = (void *)virtual_start;
554 for (i = 0; i < UPAGES; ++i) {
555 pmap_kenter_quick(virtual_start, phys_avail[0]);
556 virtual_start += PAGE_SIZE;
557 phys_avail[0] += PAGE_SIZE;
563 crashdumpmap = virtual_start;
564 virtual_start += MAXDUMPPGS * PAGE_SIZE;
567 * msgbufp maps the system message buffer
569 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
570 msgbufp = (void *)virtual_start;
571 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
572 pmap_kenter_quick(virtual_start, phys_avail[0]);
573 virtual_start += PAGE_SIZE;
574 phys_avail[0] += PAGE_SIZE;
576 msgbufinit(msgbufp, MSGBUF_SIZE);
579 * used by kern_memio for /dev/mem access
581 ptvmmap = (caddr_t)virtual_start;
582 virtual_start += PAGE_SIZE;
585 * Bootstrap the kernel_pmap
591 * Map the per-cpu globaldata for cpu #0. Allocate the space using
592 * virtual_start and phys_avail[0]
596 init_globaldata(void)
603 * Reserve enough KVA to cover possible cpus. This is a considerable
604 * amount of KVA since the privatespace structure includes two
605 * whole page table mappings.
607 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
608 CPU_prvspace = (void *)virtual_start;
609 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
612 * Allocate enough physical memory to cover the mdglobaldata
613 * portion of the space and the idle stack and map the pages
614 * into KVA. For cpu #0 only.
616 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
618 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
619 pmap_kenter_quick(va, pa);
620 phys_avail[0] += PAGE_SIZE;
622 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
624 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
625 pmap_kenter_quick(va, pa);
626 phys_avail[0] += PAGE_SIZE;
630 * Setup the %fs for cpu #0. The mycpu macro works after this
631 * point. Note that %gs is used by pthreads.
633 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
637 * Initialize very low level systems including thread0, proc0, etc.
643 struct mdglobaldata *gd;
645 gd = &CPU_prvspace[0].mdglobaldata;
646 bzero(gd, sizeof(*gd));
648 gd->mi.gd_curthread = &thread0;
649 thread0.td_gd = &gd->mi;
651 ncpus2 = 1; /* rounded down power of 2 */
652 ncpus_fit = 1; /* rounded up power of 2 */
653 /* ncpus2_mask and ncpus_fit_mask are 0 */
655 gd->mi.gd_prvspace = &CPU_prvspace[0];
656 mi_gdinit(&gd->mi, 0);
658 mi_proc0init(&gd->mi, proc0paddr);
659 lwp0.lwp_md.md_regs = &proc0_tf;
664 #if 0 /* #ifdef DDB */
666 if (boothowto & RB_KDB)
667 Debugger("Boot flags requested debugger");
670 initializecpu(); /* Initialize CPU registers */
672 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
676 * Map the message buffer
678 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
679 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
680 msgbufinit(msgbufp, MSGBUF_SIZE);
683 thread0.td_pcb_cr3 ... MMU
684 lwp0.lwp_md.md_regs = &proc0_tf;
689 * Filesystem image paths for the virtual kernel are optional.
690 * If specified they each should point to a disk image,
691 * the first of which will become the root disk.
693 * The virtual kernel caches data from our 'disk' just like a normal kernel,
694 * so we do not really want the real kernel to cache the data too. Use
695 * O_DIRECT to remove the duplication.
699 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
703 if (diskFileNum == 0)
706 for(i=0; i < diskFileNum; i++){
711 warnx("Invalid argument to '-r'");
715 if (DiskNum < VKDISK_MAX) {
717 struct vkdisk_info* info = NULL;
721 if (type == VKD_DISK)
722 fd = open(fname, O_RDWR|O_DIRECT|O_EXLOCK|O_NONBLOCK, 0644);
724 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
725 if (fd < 0 || fstat(fd, &st) < 0) {
727 fprintf(stderr, "You may already have a vkernel using this disk image!\n");
728 err(1, "Unable to open/create %s", fname);
731 /* get rid of O_NONBLOCK, keep O_DIRECT */
732 if (type == VKD_DISK)
733 fcntl(fd, F_SETFL, O_DIRECT);
735 info = &DiskInfo[DiskNum];
741 memcpy(info->fname, fname, l);
745 rootdevnames[0] = "cd9660:vcd0a";
746 else if (type == VKD_DISK)
747 rootdevnames[0] = "ufs:vkd0s0a";
752 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
760 netif_set_tapflags(int tap_unit, int f, int s)
765 bzero(&ifr, sizeof(ifr));
767 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
768 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
769 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
776 * If the flags are already set/cleared, then we return
777 * immediately to avoid extra syscalls
779 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
783 if ((flags & f) == 0)
794 * Fix up ifreq.ifr_name, since it may be trashed
795 * in previous ioctl(SIOCGIFFLAGS)
797 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
799 ifr.ifr_flags = flags & 0xffff;
800 ifr.ifr_flagshigh = flags >> 16;
801 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
802 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
810 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
812 struct ifaliasreq ifra;
813 struct sockaddr_in *in;
815 bzero(&ifra, sizeof(ifra));
816 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
819 in = (struct sockaddr_in *)&ifra.ifra_addr;
820 in->sin_family = AF_INET;
821 in->sin_len = sizeof(*in);
822 in->sin_addr.s_addr = addr;
826 in = (struct sockaddr_in *)&ifra.ifra_mask;
827 in->sin_len = sizeof(*in);
828 in->sin_addr.s_addr = mask;
831 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
832 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
840 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
845 bzero(&ifbr, sizeof(ifbr));
846 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
849 bzero(&ifd, sizeof(ifd));
850 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
851 ifd.ifd_cmd = BRDGADD;
852 ifd.ifd_len = sizeof(ifbr);
853 ifd.ifd_data = &ifbr;
855 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
857 * 'errno == EEXIST' means that the tap(4) is already
858 * a member of the bridge(4)
860 if (errno != EEXIST) {
861 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
868 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
870 /* XXX major()/minor() can't be used in vkernel */
871 #define TAPDEV_MAJOR(x) ((int)(((u_int)(x) >> 8) & 0xff))
872 #define TAPDEV_MINOR(x) ((int)((x) & 0xffff00ff))
874 #ifndef TAP_CDEV_MAJOR
875 #define TAP_CDEV_MAJOR 149
879 * Locate the first unused tap(4) device file if auto mode is requested,
880 * or open the user supplied device file, and bring up the corresponding
883 * NOTE: Only tap(4) device file is supported currently
887 netif_open_tap(const char *netif, int *tap_unit, int s)
889 char tap_dev[MAXPATHLEN];
895 if (strcmp(netif, "auto") == 0) {
900 * Find first unused tap(4) device file
903 snprintf(tap_dev, sizeof(tap_dev), "/dev/tap%d", i);
904 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
905 if (tap_fd >= 0 || errno == ENOENT)
910 warnc(lasterr, "Unable to find a free tap(4)");
915 * User supplied tap(4) device file
917 if (netif[0] == '/') /* Absolute path */
918 strlcpy(tap_dev, netif, sizeof(tap_dev));
920 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
922 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
924 warn("Unable to open %s", tap_dev);
930 * Check whether the device file is a tap(4)
933 if (fstat(tap_fd, &st) == 0 && S_ISCHR(st.st_mode) &&
934 TAPDEV_MAJOR(st.st_rdev) == TAP_CDEV_MAJOR) {
935 *tap_unit = TAPDEV_MINOR(st.st_rdev);
938 * Bring up the corresponding tap(4) interface
940 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
943 warnx("%s is not a tap(4) device", tap_dev);
959 * Following syntax is supported,
960 * 1) x.x.x.x tap(4)'s address is x.x.x.x
962 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
963 * tap(4)'s netmask len is z
965 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
966 * pseudo netif's address is y.y.y.y
968 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
969 * pseudo netif's address is y.y.y.y
970 * tap(4) and pseudo netif's netmask len are z
972 * 5) bridgeX tap(4) will be added to bridgeX
974 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
975 * pseudo netif's address is y.y.y.y
977 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
978 * pseudo netif's address is y.y.y.y
979 * pseudo netif's netmask len is z
983 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
985 in_addr_t tap_addr, netmask, netif_addr;
987 char *tok, *masklen_str, *ifbridge;
992 tok = strtok(NULL, ":/");
995 * Nothing special, simply use tap(4) as backend
1000 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1002 * tap(4)'s address is supplied
1007 * If there is next token, then it may be pseudo
1008 * netif's address or netmask len for tap(4)
1010 next_netif_addr = 0;
1013 * Not tap(4)'s address, assume it as a bridge(4)
1020 * If there is next token, then it must be pseudo
1023 next_netif_addr = 1;
1026 netmask = netif_addr = 0;
1028 tok = strtok(NULL, ":/");
1032 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1033 if (next_netif_addr) {
1034 warnx("Invalid pseudo netif address: %s", tok);
1040 * Current token is not address, then it must be netmask len
1045 * Current token is pseudo netif address, if there is next token
1046 * it must be netmask len
1048 masklen_str = strtok(NULL, "/");
1051 /* Calculate netmask */
1052 if (masklen_str != NULL) {
1055 masklen = strtoul(masklen_str, NULL, 10);
1056 if (masklen < 32 && masklen > 0) {
1057 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1060 warnx("Invalid netmask len: %lu", masklen);
1065 /* Make sure there is no more token left */
1066 if (strtok(NULL, ":/") != NULL) {
1067 warnx("Invalid argument to '-I'");
1072 if (ifbridge == NULL) {
1073 /* Set tap(4) address/netmask */
1074 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1077 /* Tie tap(4) to bridge(4) */
1078 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1088 * NetifInfo[] will be filled for pseudo netif initialization.
1089 * NetifNum will be bumped to reflect the number of valid entries
1094 init_netif(char *netifExp[], int netifExpNum)
1098 if (netifExpNum == 0)
1101 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1105 for (i = 0; i < netifExpNum; ++i) {
1106 struct vknetif_info *info;
1107 in_addr_t netif_addr, netif_mask;
1108 int tap_fd, tap_unit;
1111 netif = strtok(netifExp[i], ":");
1112 if (netif == NULL) {
1113 warnx("Invalid argument to '-I'");
1118 * Open tap(4) device file and bring up the
1119 * corresponding interface
1121 tap_fd = netif_open_tap(netif, &tap_unit, s);
1126 * Initialize tap(4) and get address/netmask
1129 * NB: Rest part of netifExp[i] is passed
1130 * to netif_init_tap() implicitly.
1132 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1134 * NB: Closing tap(4) device file will bring
1135 * down the corresponding interface
1141 info = &NetifInfo[NetifNum];
1142 info->tap_fd = tap_fd;
1143 info->tap_unit = tap_unit;
1144 info->netif_addr = netif_addr;
1145 info->netif_mask = netif_mask;
1148 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1161 if (pid_file != NULL) {
1163 fp = fopen(pid_file, "w");
1166 fprintf(fp, "%ld\n", (long)self);
1170 perror("Warning: couldn't open pidfile");
1179 if (pid_file != NULL) {
1180 if ( unlink(pid_file) != 0 )
1181 perror("Warning: couldn't remove pidfile");
1187 usage(const char *ctl, ...)
1192 vfprintf(stderr, ctl, va);
1194 fprintf(stderr, "\n");
1201 kprintf("cpu reset, rebooting vkernel\n");
1204 execv(save_av[0], save_av);
1210 kprintf("cpu halt, exiting vkernel\n");
1218 switch(lwp_cpu_lock) {
1221 kprintf("Locking CPU%d to real cpu %d\n",
1223 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1225 if (next_cpu >= real_ncpus)
1228 case LCL_SINGLE_CPU:
1230 kprintf("Locking CPU%d to real cpu %d\n",
1232 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1235 /* do not map virtual cpus to real cpus */