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 1.56 2008/05/27 07:48:00 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>
54 #include <vm/vm_page.h>
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>
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>
80 vm_paddr_t phys_avail[16];
82 vm_paddr_t Maxmem_bytes;
85 struct vkdisk_info DiskInfo[VKDISK_MAX];
87 struct vknetif_info NetifInfo[VKNETIF_MAX];
93 vm_offset_t virtual_start;
94 vm_offset_t virtual_end;
95 vm_offset_t virtual2_start;
96 vm_offset_t virtual2_end;
97 vm_offset_t kernel_vm_end;
98 vm_offset_t crashdumpmap;
99 vm_offset_t clean_sva;
100 vm_offset_t clean_eva;
101 struct msgbuf *msgbufp;
104 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
105 void *dmap_min_address;
106 u_int cpu_feature; /* XXX */
108 int64_t tsc_frequency;
109 int optcpus; /* number of cpus - see mp_start() */
110 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
111 int real_ncpus; /* number of real CPUs */
112 int next_cpu; /* next real CPU to lock a virtual CPU to */
114 struct privatespace *CPU_prvspace;
116 static struct trapframe proc0_tf;
117 static void *proc0paddr;
119 static void init_sys_memory(char *imageFile);
120 static void init_kern_memory(void);
121 static void init_globaldata(void);
122 static void init_vkernel(void);
123 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
124 static void init_netif(char *netifExp[], int netifFileNum);
125 static void writepid( void );
126 static void cleanpid( void );
127 static int unix_connect(const char *path);
128 static void usage(const char *ctl, ...);
131 static char **save_av;
134 * Kernel startup for virtual kernels - standard main()
137 main(int ac, char **av)
139 char *memImageFile = NULL;
140 char *netifFile[VKNETIF_MAX];
141 char *diskFile[VKDISK_MAX];
142 char *cdFile[VKDISK_MAX];
145 int netifFileNum = 0;
148 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
154 int real_vkernel_enable;
164 kernel_mem_readonly = 1;
168 lwp_cpu_lock = LCL_NONE;
170 real_vkernel_enable = 0;
171 vsize = sizeof(real_vkernel_enable);
172 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
174 if (real_vkernel_enable == 0) {
175 errx(1, "vm.vkernel_enable is 0, must be set "
176 "to 1 to execute a vkernel!");
180 vsize = sizeof(real_ncpus);
181 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
183 while ((c = getopt(ac, av, "c:svl:m:n:r:e:i:p:I:U")) != -1) {
187 * name=value:name=value:name=value...
190 * Allow values to be quoted but note that shells
191 * may remove the quotes, so using this feature
192 * to embed colons may require a backslash.
196 kern_envp = malloc(n + 2);
197 for (i = j = 0; i < n; ++i) {
198 if (optarg[i] == '"')
200 else if (optarg[i] == '\'')
202 else if (isq == 0 && optarg[i] == ':')
205 kern_envp[j++] = optarg[i];
211 boothowto |= RB_SINGLE;
217 memImageFile = optarg;
220 if (netifFileNum < VKNETIF_MAX)
221 netifFile[netifFileNum++] = strdup(optarg);
226 if (diskFileNum + cdFileNum < VKDISK_MAX)
227 diskFile[diskFileNum++] = strdup(optarg);
232 if (diskFileNum + cdFileNum < VKDISK_MAX)
233 cdFile[cdFileNum++] = strdup(optarg);
236 Maxmem_bytes = strtoull(optarg, &suffix, 0);
253 usage("Bad maxmem option");
261 if (strncmp("map", optarg, 3) == 0) {
262 lwp_cpu_lock = LCL_PER_CPU;
263 if (optarg[3] == ',') {
264 next_cpu = strtol(optarg+4, &endp, 0);
266 usage("Bad target CPU number at '%s'", endp);
270 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
271 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
272 } else if (strncmp("any", optarg, 3) == 0) {
273 lwp_cpu_lock = LCL_NONE;
275 lwp_cpu_lock = LCL_SINGLE_CPU;
276 next_cpu = strtol(optarg, &endp, 0);
278 usage("Bad target CPU number at '%s'", endp);
279 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
280 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
285 * This value is set up by mp_start(), don't just
289 optcpus = strtol(optarg, NULL, 0);
290 if (optcpus < 1 || optcpus > MAXCPU)
291 usage("Bad ncpus, valid range is 1-%d", MAXCPU);
293 if (strtol(optarg, NULL, 0) != 1) {
294 usage("You built a UP vkernel, only 1 cpu!");
303 kernel_mem_readonly = 0;
310 init_sys_memory(memImageFile);
320 vsize = sizeof(tsc_present);
321 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
322 vsize = sizeof(tsc_frequency);
323 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
325 cpu_feature |= CPUID_TSC;
330 vsize = sizeof(supports_sse);
332 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
333 init_fpu(supports_sse);
335 cpu_feature |= CPUID_SSE | CPUID_FXSR;
338 * We boot from the first installed disk.
340 if (bootOnDisk == 1) {
341 init_disk(diskFile, diskFileNum, VKD_DISK);
342 init_disk(cdFile, cdFileNum, VKD_CD);
344 init_disk(cdFile, cdFileNum, VKD_CD);
345 init_disk(diskFile, diskFileNum, VKD_DISK);
347 init_netif(netifFile, netifFileNum);
355 * Initialize system memory. This is the virtual kernel's 'RAM'.
359 init_sys_memory(char *imageFile)
366 * Figure out the system memory image size. If an image file was
367 * specified and -m was not specified, use the image file's size.
370 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
371 Maxmem_bytes = (vm_paddr_t)st.st_size;
372 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
374 err(1, "Cannot create new memory file %s unless "
375 "system memory size is specified with -m",
381 * Maxmem must be known at this time
383 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
384 err(1, "Bad maxmem specification: 32MB minimum, "
385 "multiples of %dMB only",
386 SEG_SIZE / 1024 / 1024);
391 * Generate an image file name if necessary, then open/create the
392 * file exclusively locked. Do not allow multiple virtual kernels
393 * to use the same image file.
395 if (imageFile == NULL) {
396 for (i = 0; i < 1000000; ++i) {
397 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
399 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
400 if (fd < 0 && errno == EWOULDBLOCK) {
407 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
409 fprintf(stderr, "Using memory file: %s\n", imageFile);
410 if (fd < 0 || fstat(fd, &st) < 0) {
411 err(1, "Unable to open/create %s", imageFile);
416 * Truncate or extend the file as necessary. Clean out the contents
417 * of the file, we want it to be full of holes so we don't waste
418 * time reading in data from an old file that we no longer care
422 ftruncate(fd, Maxmem_bytes);
425 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
430 * Initialize kernel memory. This reserves kernel virtual memory by using
436 init_kern_memory(void)
441 char *topofstack = &dummy;
446 * Memory map our kernel virtual memory space. Note that the
447 * kernel image itself is not made part of this memory for the
450 * The memory map must be segment-aligned so we can properly
453 * If the system kernel has a different MAXDSIZ, it might not
454 * be possible to map kernel memory in its prefered location.
455 * Try a number of different locations.
457 try = (void *)(512UL << 30);
459 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
460 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
461 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
462 MemImageFd, (off_t)try);
465 if (base != MAP_FAILED)
466 munmap(base, KERNEL_KVA_SIZE);
467 try = (char *)try + (512UL << 30);
470 err(1, "Unable to mmap() kernel virtual memory!");
473 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
474 KvaStart = (vm_offset_t)base;
475 KvaSize = KERNEL_KVA_SIZE;
476 KvaEnd = KvaStart + KvaSize;
478 /* cannot use kprintf yet */
479 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
482 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
483 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
485 if (dmap_min_address == MAP_FAILED) {
486 err(1, "Unable to mmap() kernel DMAP region!");
491 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
493 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
494 0 | VPTE_R | VPTE_W | VPTE_V);
497 * phys_avail[] represents unallocated physical memory. MI code
498 * will use phys_avail[] to create the vm_page array.
500 phys_avail[0] = (vm_paddr_t)firstfree;
501 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
502 phys_avail[1] = Maxmem_bytes;
506 * (virtual_start, virtual_end) represent unallocated kernel virtual
507 * memory. MI code will create kernel_map using these parameters.
509 virtual_start = KvaStart + (long)firstfree;
510 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
511 virtual_end = KvaStart + KERNEL_KVA_SIZE;
515 * pmap_growkernel() will set the correct value.
520 * Allocate space for process 0's UAREA.
522 proc0paddr = (void *)virtual_start;
523 for (i = 0; i < UPAGES; ++i) {
524 pmap_kenter_quick(virtual_start, phys_avail[0]);
525 virtual_start += PAGE_SIZE;
526 phys_avail[0] += PAGE_SIZE;
532 crashdumpmap = virtual_start;
533 virtual_start += MAXDUMPPGS * PAGE_SIZE;
536 * msgbufp maps the system message buffer
538 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
539 msgbufp = (void *)virtual_start;
540 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
541 pmap_kenter_quick(virtual_start, phys_avail[0]);
542 virtual_start += PAGE_SIZE;
543 phys_avail[0] += PAGE_SIZE;
545 msgbufinit(msgbufp, MSGBUF_SIZE);
548 * used by kern_memio for /dev/mem access
550 ptvmmap = (caddr_t)virtual_start;
551 virtual_start += PAGE_SIZE;
554 * Bootstrap the kernel_pmap
562 * Map the per-cpu globaldata for cpu #0. Allocate the space using
563 * virtual_start and phys_avail[0]
567 init_globaldata(void)
574 * Reserve enough KVA to cover possible cpus. This is a considerable
575 * amount of KVA since the privatespace structure includes two
576 * whole page table mappings.
578 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
579 CPU_prvspace = (void *)virtual_start;
580 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
583 * Allocate enough physical memory to cover the mdglobaldata
584 * portion of the space and the idle stack and map the pages
585 * into KVA. For cpu #0 only.
587 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
589 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
590 pmap_kenter_quick(va, pa);
591 phys_avail[0] += PAGE_SIZE;
593 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
595 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
596 pmap_kenter_quick(va, pa);
597 phys_avail[0] += PAGE_SIZE;
601 * Setup the %gs for cpu #0. The mycpu macro works after this
602 * point. Note that %fs is used by pthreads.
604 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
608 * Initialize very low level systems including thread0, proc0, etc.
614 struct mdglobaldata *gd;
616 gd = &CPU_prvspace[0].mdglobaldata;
617 bzero(gd, sizeof(*gd));
619 gd->mi.gd_curthread = &thread0;
620 thread0.td_gd = &gd->mi;
622 ncpus2 = 1; /* rounded down power of 2 */
623 ncpus_fit = 1; /* rounded up power of 2 */
624 /* ncpus2_mask and ncpus_fit_mask are 0 */
626 gd->mi.gd_prvspace = &CPU_prvspace[0];
627 mi_gdinit(&gd->mi, 0);
629 mi_proc0init(&gd->mi, proc0paddr);
630 lwp0.lwp_md.md_regs = &proc0_tf;
635 #if 0 /* #ifdef DDB */
637 if (boothowto & RB_KDB)
638 Debugger("Boot flags requested debugger");
642 initializecpu(); /* Initialize CPU registers */
644 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
648 * Map the message buffer
650 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
651 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
652 msgbufinit(msgbufp, MSGBUF_SIZE);
655 thread0.td_pcb_cr3 ... MMU
656 lwp0.lwp_md.md_regs = &proc0_tf;
661 * Filesystem image paths for the virtual kernel are optional.
662 * If specified they each should point to a disk image,
663 * the first of which will become the root disk.
665 * The virtual kernel caches data from our 'disk' just like a normal kernel,
666 * so we do not really want the real kernel to cache the data too. Use
667 * O_DIRECT to remove the duplication.
671 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
675 if (diskFileNum == 0)
678 for(i=0; i < diskFileNum; i++){
683 warnx("Invalid argument to '-r'");
687 if (DiskNum < VKDISK_MAX) {
689 struct vkdisk_info* info = NULL;
693 if (type == VKD_DISK)
694 fd = open(fname, O_RDWR|O_DIRECT, 0644);
696 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
697 if (fd < 0 || fstat(fd, &st) < 0) {
698 err(1, "Unable to open/create %s", fname);
701 if (S_ISREG(st.st_mode)) {
702 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
703 errx(1, "Disk image %s is already "
709 info = &DiskInfo[DiskNum];
715 memcpy(info->fname, fname, l);
718 if (type == VKD_CD) {
719 rootdevnames[0] = "cd9660:vcd0a";
720 } else if (type == VKD_DISK) {
721 rootdevnames[0] = "ufs:vkd0s0a";
722 rootdevnames[1] = "ufs:vkd0s1a";
728 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
736 netif_set_tapflags(int tap_unit, int f, int s)
741 bzero(&ifr, sizeof(ifr));
743 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
744 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
745 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
752 * If the flags are already set/cleared, then we return
753 * immediately to avoid extra syscalls
755 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
759 if ((flags & f) == 0)
770 * Fix up ifreq.ifr_name, since it may be trashed
771 * in previous ioctl(SIOCGIFFLAGS)
773 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
775 ifr.ifr_flags = flags & 0xffff;
776 ifr.ifr_flagshigh = flags >> 16;
777 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
778 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
786 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
788 struct ifaliasreq ifra;
789 struct sockaddr_in *in;
791 bzero(&ifra, sizeof(ifra));
792 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
795 in = (struct sockaddr_in *)&ifra.ifra_addr;
796 in->sin_family = AF_INET;
797 in->sin_len = sizeof(*in);
798 in->sin_addr.s_addr = addr;
802 in = (struct sockaddr_in *)&ifra.ifra_mask;
803 in->sin_len = sizeof(*in);
804 in->sin_addr.s_addr = mask;
807 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
808 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
816 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
821 bzero(&ifbr, sizeof(ifbr));
822 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
825 bzero(&ifd, sizeof(ifd));
826 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
827 ifd.ifd_cmd = BRDGADD;
828 ifd.ifd_len = sizeof(ifbr);
829 ifd.ifd_data = &ifbr;
831 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
833 * 'errno == EEXIST' means that the tap(4) is already
834 * a member of the bridge(4)
836 if (errno != EEXIST) {
837 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
844 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
847 * Locate the first unused tap(4) device file if auto mode is requested,
848 * or open the user supplied device file, and bring up the corresponding
851 * NOTE: Only tap(4) device file is supported currently
855 netif_open_tap(const char *netif, int *tap_unit, int s)
857 char tap_dev[MAXPATHLEN];
864 if (strcmp(netif, "auto") == 0) {
866 * Find first unused tap(4) device file
868 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
870 warnc(errno, "Unable to find a free tap(4)");
875 * User supplied tap(4) device file or unix socket.
877 if (netif[0] == '/') /* Absolute path */
878 strlcpy(tap_dev, netif, sizeof(tap_dev));
880 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
882 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
885 * If we cannot open normally try to connect to it.
888 tap_fd = unix_connect(tap_dev);
891 warn("Unable to open %s", tap_dev);
897 * Check whether the device file is a tap(4)
899 if (fstat(tap_fd, &st) < 0) {
901 } else if (S_ISCHR(st.st_mode)) {
902 dname = fdevname(tap_fd);
904 dname = strstr(dname, "tap");
907 * Bring up the corresponding tap(4) interface
909 *tap_unit = strtol(dname + 3, NULL, 10);
910 printf("TAP UNIT %d\n", *tap_unit);
911 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
918 } else if (S_ISSOCK(st.st_mode)) {
920 * Special socket connection (typically to vknet). We
921 * do not have to do anything.
929 warnx("%s is not a tap(4) device or socket", tap_dev);
938 unix_connect(const char *path)
940 struct sockaddr_un sunx;
946 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
947 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
948 ++len; /* include nul */
949 sunx.sun_family = AF_UNIX;
952 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
955 if (connect(net_fd, (void *)&sunx, len) < 0) {
959 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
960 if (fstat(net_fd, &st) == 0)
961 printf("Network socket buffer: %d bytes\n", st.st_blksize);
962 fcntl(net_fd, F_SETFL, O_NONBLOCK);
971 * Following syntax is supported,
972 * 1) x.x.x.x tap(4)'s address is x.x.x.x
974 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
975 * tap(4)'s netmask len is z
977 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
978 * pseudo netif's address is y.y.y.y
980 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
981 * pseudo netif's address is y.y.y.y
982 * tap(4) and pseudo netif's netmask len are z
984 * 5) bridgeX tap(4) will be added to bridgeX
986 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
987 * pseudo netif's address is y.y.y.y
989 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
990 * pseudo netif's address is y.y.y.y
991 * pseudo netif's netmask len is z
995 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
997 in_addr_t tap_addr, netmask, netif_addr;
999 char *tok, *masklen_str, *ifbridge;
1004 tok = strtok(NULL, ":/");
1007 * Nothing special, simply use tap(4) as backend
1012 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1014 * tap(4)'s address is supplied
1019 * If there is next token, then it may be pseudo
1020 * netif's address or netmask len for tap(4)
1022 next_netif_addr = 0;
1025 * Not tap(4)'s address, assume it as a bridge(4)
1032 * If there is next token, then it must be pseudo
1035 next_netif_addr = 1;
1038 netmask = netif_addr = 0;
1040 tok = strtok(NULL, ":/");
1044 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1045 if (next_netif_addr) {
1046 warnx("Invalid pseudo netif address: %s", tok);
1052 * Current token is not address, then it must be netmask len
1057 * Current token is pseudo netif address, if there is next token
1058 * it must be netmask len
1060 masklen_str = strtok(NULL, "/");
1063 /* Calculate netmask */
1064 if (masklen_str != NULL) {
1067 masklen = strtoul(masklen_str, NULL, 10);
1068 if (masklen < 32 && masklen > 0) {
1069 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1072 warnx("Invalid netmask len: %lu", masklen);
1077 /* Make sure there is no more token left */
1078 if (strtok(NULL, ":/") != NULL) {
1079 warnx("Invalid argument to '-I'");
1086 } else if (ifbridge == NULL) {
1087 /* Set tap(4) address/netmask */
1088 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1091 /* Tie tap(4) to bridge(4) */
1092 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1102 * NetifInfo[] will be filled for pseudo netif initialization.
1103 * NetifNum will be bumped to reflect the number of valid entries
1108 init_netif(char *netifExp[], int netifExpNum)
1112 if (netifExpNum == 0)
1115 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1119 for (i = 0; i < netifExpNum; ++i) {
1120 struct vknetif_info *info;
1121 in_addr_t netif_addr, netif_mask;
1122 int tap_fd, tap_unit;
1125 netif = strtok(netifExp[i], ":");
1126 if (netif == NULL) {
1127 warnx("Invalid argument to '-I'");
1132 * Open tap(4) device file and bring up the
1133 * corresponding interface
1135 tap_fd = netif_open_tap(netif, &tap_unit, s);
1140 * Initialize tap(4) and get address/netmask
1143 * NB: Rest part of netifExp[i] is passed
1144 * to netif_init_tap() implicitly.
1146 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1148 * NB: Closing tap(4) device file will bring
1149 * down the corresponding interface
1155 info = &NetifInfo[NetifNum];
1156 info->tap_fd = tap_fd;
1157 info->tap_unit = tap_unit;
1158 info->netif_addr = netif_addr;
1159 info->netif_mask = netif_mask;
1162 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1175 if (pid_file != NULL) {
1177 fp = fopen(pid_file, "w");
1180 fprintf(fp, "%ld\n", (long)self);
1184 perror("Warning: couldn't open pidfile");
1193 if (pid_file != NULL) {
1194 if ( unlink(pid_file) != 0 )
1195 perror("Warning: couldn't remove pidfile");
1201 usage(const char *ctl, ...)
1206 vfprintf(stderr, ctl, va);
1208 fprintf(stderr, "\n");
1215 kprintf("cpu reset, rebooting vkernel\n");
1218 execv(save_av[0], save_av);
1224 kprintf("cpu halt, exiting vkernel\n");
1232 switch(lwp_cpu_lock) {
1235 kprintf("Locking CPU%d to real cpu %d\n",
1237 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1239 if (next_cpu >= real_ncpus)
1242 case LCL_SINGLE_CPU:
1244 kprintf("Locking CPU%d to real cpu %d\n",
1246 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1249 /* do not map virtual cpus to real cpus */