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;
84 struct vkdisk_info DiskInfo[VKDISK_MAX];
86 struct vknetif_info NetifInfo[VKNETIF_MAX];
92 vm_offset_t virtual_start;
93 vm_offset_t virtual_end;
94 vm_offset_t kernel_vm_end;
95 vm_offset_t crashdumpmap;
96 vm_offset_t clean_sva;
97 vm_offset_t clean_eva;
98 struct msgbuf *msgbufp;
101 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
102 u_int cpu_feature; /* XXX */
104 int64_t tsc_frequency;
105 int optcpus; /* number of cpus - see mp_start() */
106 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
107 int real_ncpus; /* number of real CPUs */
108 int next_cpu; /* next real CPU to lock a virtual CPU to */
110 struct privatespace *CPU_prvspace;
112 static struct trapframe proc0_tf;
113 static void *proc0paddr;
115 static void init_sys_memory(char *imageFile);
116 static void init_kern_memory(void);
117 static void init_globaldata(void);
118 static void init_vkernel(void);
119 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
120 static void init_netif(char *netifExp[], int netifFileNum);
121 static void writepid( void );
122 static void cleanpid( void );
123 static int unix_connect(const char *path);
124 static void usage(const char *ctl, ...);
127 static char **save_av;
130 * Kernel startup for virtual kernels - standard main()
133 main(int ac, char **av)
135 char *memImageFile = NULL;
136 char *netifFile[VKNETIF_MAX];
137 char *diskFile[VKDISK_MAX];
138 char *cdFile[VKDISK_MAX];
141 int netifFileNum = 0;
144 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
148 int real_vkernel_enable;
158 kernel_mem_readonly = 1;
162 lwp_cpu_lock = LCL_NONE;
164 real_vkernel_enable = 0;
165 vsize = sizeof(real_vkernel_enable);
166 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
168 if (real_vkernel_enable == 0) {
169 errx(1, "vm.vkernel_enable is 0, must be set "
170 "to 1 to execute a vkernel!");
174 vsize = sizeof(real_ncpus);
175 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
177 while ((c = getopt(ac, av, "c:svl:m:n:r:e:i:p:I:U")) != -1) {
181 * name=value:name=value:name=value...
184 kern_envp = malloc(n + 2);
185 for (i = 0; i < n; ++i) {
186 if (optarg[i] == ':')
189 kern_envp[i] = optarg[i];
195 boothowto |= RB_SINGLE;
201 memImageFile = optarg;
204 if (netifFileNum < VKNETIF_MAX)
205 netifFile[netifFileNum++] = strdup(optarg);
210 if (diskFileNum + cdFileNum < VKDISK_MAX)
211 diskFile[diskFileNum++] = strdup(optarg);
216 if (diskFileNum + cdFileNum < VKDISK_MAX)
217 cdFile[cdFileNum++] = strdup(optarg);
220 Maxmem_bytes = strtoull(optarg, &suffix, 0);
237 usage("Bad maxmem option");
245 if (strncmp("map", optarg, 3) == 0) {
246 lwp_cpu_lock = LCL_PER_CPU;
247 if (optarg[3] == ',') {
248 next_cpu = strtol(optarg+4, &endp, 0);
250 usage("Bad target CPU number at '%s'", endp);
254 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
255 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
256 } else if (strncmp("any", optarg, 3) == 0) {
257 lwp_cpu_lock = LCL_NONE;
259 lwp_cpu_lock = LCL_SINGLE_CPU;
260 next_cpu = strtol(optarg, &endp, 0);
262 usage("Bad target CPU number at '%s'", endp);
263 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
264 usage("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
269 * This value is set up by mp_start(), don't just
273 optcpus = strtol(optarg, NULL, 0);
274 if (optcpus < 1 || optcpus > MAXCPU)
275 usage("Bad ncpus, valid range is 1-%d", MAXCPU);
277 if (strtol(optarg, NULL, 0) != 1) {
278 usage("You built a UP vkernel, only 1 cpu!");
287 kernel_mem_readonly = 0;
294 init_sys_memory(memImageFile);
304 vsize = sizeof(tsc_present);
305 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
306 vsize = sizeof(tsc_frequency);
307 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
309 cpu_feature |= CPUID_TSC;
314 vsize = sizeof(supports_sse);
316 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
317 init_fpu(supports_sse);
319 cpu_feature |= CPUID_SSE | CPUID_FXSR;
322 * We boot from the first installed disk.
324 if (bootOnDisk == 1) {
325 init_disk(diskFile, diskFileNum, VKD_DISK);
326 init_disk(cdFile, cdFileNum, VKD_CD);
328 init_disk(cdFile, cdFileNum, VKD_CD);
329 init_disk(diskFile, diskFileNum, VKD_DISK);
331 init_netif(netifFile, netifFileNum);
339 * Initialize system memory. This is the virtual kernel's 'RAM'.
343 init_sys_memory(char *imageFile)
350 * Figure out the system memory image size. If an image file was
351 * specified and -m was not specified, use the image file's size.
354 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
355 Maxmem_bytes = (vm_paddr_t)st.st_size;
356 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
358 err(1, "Cannot create new memory file %s unless "
359 "system memory size is specified with -m",
365 * Maxmem must be known at this time
367 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
368 err(1, "Bad maxmem specification: 32MB minimum, "
369 "multiples of %dMB only",
370 SEG_SIZE / 1024 / 1024);
375 * Generate an image file name if necessary, then open/create the
376 * file exclusively locked. Do not allow multiple virtual kernels
377 * to use the same image file.
379 if (imageFile == NULL) {
380 for (i = 0; i < 1000000; ++i) {
381 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
383 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
384 if (fd < 0 && errno == EWOULDBLOCK) {
391 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
393 printf("Using memory file: %s\n", imageFile);
394 if (fd < 0 || fstat(fd, &st) < 0) {
395 err(1, "Unable to open/create %s", imageFile);
400 * Truncate or extend the file as necessary.
402 if (st.st_size > Maxmem_bytes) {
403 ftruncate(fd, Maxmem_bytes);
404 } else if (st.st_size < Maxmem_bytes) {
406 off_t off = st.st_size & ~SEG_MASK;
408 kprintf("%s: Reserving blocks for memory image\n", imageFile);
409 zmem = malloc(SEG_SIZE);
410 bzero(zmem, SEG_SIZE);
411 lseek(fd, off, SEEK_SET);
412 while (off < Maxmem_bytes) {
413 if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
414 err(1, "Unable to reserve blocks for memory image");
420 err(1, "Unable to reserve blocks for memory image");
424 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
428 * Initialize kernel memory. This reserves kernel virtual memory by using
434 init_kern_memory(void)
440 char *topofstack = &dummy;
445 * Memory map our kernel virtual memory space. Note that the
446 * kernel image itself is not made part of this memory for the
449 * The memory map must be segment-aligned so we can properly
452 * If the system kernel has a different MAXDSIZ, it might not
453 * be possible to map kernel memory in its prefered location.
454 * Try a number of different locations.
456 try = (void *)0x40000000;
458 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
459 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
460 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
464 if (base != MAP_FAILED)
465 munmap(base, KERNEL_KVA_SIZE);
466 try = (char *)try + 0x10000000;
469 err(1, "Unable to mmap() kernel virtual memory!");
472 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
473 KvaStart = (vm_offset_t)base;
474 KvaSize = KERNEL_KVA_SIZE;
475 KvaEnd = KvaStart + KvaSize;
476 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
479 * Create a top-level page table self-mapping itself.
481 * Initialize the page directory at physical page index 0 to point
482 * to an array of page table pages starting at physical page index 1
484 lseek(MemImageFd, 0L, 0);
485 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
486 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
487 write(MemImageFd, &pte, sizeof(pte));
491 * Initialize the PTEs in the page table pages required to map the
492 * page table itself. This includes mapping the page directory page
493 * at the base so we go one more loop then normal.
495 lseek(MemImageFd, PAGE_SIZE, 0);
496 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
497 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
498 write(MemImageFd, &pte, sizeof(pte));
502 * Initialize remaining PTEs to 0. We may be reusing a memory image
503 * file. This is approximately a megabyte.
505 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
506 zero = malloc(PAGE_SIZE);
507 bzero(zero, PAGE_SIZE);
509 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
510 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
515 * Enable the page table and calculate pointers to our self-map
516 * for easy kernel page table manipulation.
518 * KernelPTA must be offset so we can do direct VA translations
520 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
521 0 | VPTE_R | VPTE_W | VPTE_V);
522 KernelPTD = (vpte_t *)base; /* pg directory */
523 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
524 KernelPTA -= KvaStart >> PAGE_SHIFT;
527 * phys_avail[] represents unallocated physical memory. MI code
528 * will use phys_avail[] to create the vm_page array.
530 phys_avail[0] = PAGE_SIZE +
531 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
532 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
533 phys_avail[1] = Maxmem_bytes;
536 * (virtual_start, virtual_end) represent unallocated kernel virtual
537 * memory. MI code will create kernel_map using these parameters.
539 virtual_start = KvaStart + PAGE_SIZE +
540 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
541 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
542 virtual_end = KvaStart + KERNEL_KVA_SIZE;
545 * kernel_vm_end could be set to virtual_end but we want some
546 * indication of how much of the kernel_map we've used, so
547 * set it low and let pmap_growkernel increase it even though we
548 * don't need to create any new page table pages.
550 kernel_vm_end = virtual_start;
553 * Allocate space for process 0's UAREA.
555 proc0paddr = (void *)virtual_start;
556 for (i = 0; i < UPAGES; ++i) {
557 pmap_kenter_quick(virtual_start, phys_avail[0]);
558 virtual_start += PAGE_SIZE;
559 phys_avail[0] += PAGE_SIZE;
565 crashdumpmap = virtual_start;
566 virtual_start += MAXDUMPPGS * PAGE_SIZE;
569 * msgbufp maps the system message buffer
571 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
572 msgbufp = (void *)virtual_start;
573 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
574 pmap_kenter_quick(virtual_start, phys_avail[0]);
575 virtual_start += PAGE_SIZE;
576 phys_avail[0] += PAGE_SIZE;
578 msgbufinit(msgbufp, MSGBUF_SIZE);
581 * used by kern_memio for /dev/mem access
583 ptvmmap = (caddr_t)virtual_start;
584 virtual_start += PAGE_SIZE;
587 * Bootstrap the kernel_pmap
593 * Map the per-cpu globaldata for cpu #0. Allocate the space using
594 * virtual_start and phys_avail[0]
598 init_globaldata(void)
605 * Reserve enough KVA to cover possible cpus. This is a considerable
606 * amount of KVA since the privatespace structure includes two
607 * whole page table mappings.
609 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
610 CPU_prvspace = (void *)virtual_start;
611 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
614 * Allocate enough physical memory to cover the mdglobaldata
615 * portion of the space and the idle stack and map the pages
616 * into KVA. For cpu #0 only.
618 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
620 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
621 pmap_kenter_quick(va, pa);
622 phys_avail[0] += PAGE_SIZE;
624 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
626 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
627 pmap_kenter_quick(va, pa);
628 phys_avail[0] += PAGE_SIZE;
632 * Setup the %fs for cpu #0. The mycpu macro works after this
633 * point. Note that %gs is used by pthreads.
635 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
639 * Initialize very low level systems including thread0, proc0, etc.
645 struct mdglobaldata *gd;
647 gd = &CPU_prvspace[0].mdglobaldata;
648 bzero(gd, sizeof(*gd));
650 gd->mi.gd_curthread = &thread0;
651 thread0.td_gd = &gd->mi;
653 ncpus2 = 1; /* rounded down power of 2 */
654 ncpus_fit = 1; /* rounded up power of 2 */
655 /* ncpus2_mask and ncpus_fit_mask are 0 */
657 gd->mi.gd_prvspace = &CPU_prvspace[0];
658 mi_gdinit(&gd->mi, 0);
660 mi_proc0init(&gd->mi, proc0paddr);
661 lwp0.lwp_md.md_regs = &proc0_tf;
666 #if 0 /* #ifdef DDB */
668 if (boothowto & RB_KDB)
669 Debugger("Boot flags requested debugger");
673 initializecpu(); /* Initialize CPU registers */
675 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
679 * Map the message buffer
681 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
682 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
683 msgbufinit(msgbufp, MSGBUF_SIZE);
686 thread0.td_pcb_cr3 ... MMU
687 lwp0.lwp_md.md_regs = &proc0_tf;
692 * Filesystem image paths for the virtual kernel are optional.
693 * If specified they each should point to a disk image,
694 * the first of which will become the root disk.
696 * The virtual kernel caches data from our 'disk' just like a normal kernel,
697 * so we do not really want the real kernel to cache the data too. Use
698 * O_DIRECT to remove the duplication.
702 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
706 if (diskFileNum == 0)
709 for(i=0; i < diskFileNum; i++){
714 warnx("Invalid argument to '-r'");
718 if (DiskNum < VKDISK_MAX) {
720 struct vkdisk_info* info = NULL;
724 if (type == VKD_DISK)
725 fd = open(fname, O_RDWR|O_DIRECT|O_EXLOCK|O_NONBLOCK, 0644);
727 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
728 if (fd < 0 || fstat(fd, &st) < 0) {
730 fprintf(stderr, "You may already have a vkernel using this disk image!\n");
731 err(1, "Unable to open/create %s", fname);
734 /* get rid of O_NONBLOCK, keep O_DIRECT */
735 if (type == VKD_DISK)
736 fcntl(fd, F_SETFL, O_DIRECT);
738 info = &DiskInfo[DiskNum];
744 memcpy(info->fname, fname, l);
748 rootdevnames[0] = "cd9660:vcd0a";
749 else if (type == VKD_DISK)
750 rootdevnames[0] = "ufs:vkd0s0a";
755 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
763 netif_set_tapflags(int tap_unit, int f, int s)
768 bzero(&ifr, sizeof(ifr));
770 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
771 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
772 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
779 * If the flags are already set/cleared, then we return
780 * immediately to avoid extra syscalls
782 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
786 if ((flags & f) == 0)
797 * Fix up ifreq.ifr_name, since it may be trashed
798 * in previous ioctl(SIOCGIFFLAGS)
800 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
802 ifr.ifr_flags = flags & 0xffff;
803 ifr.ifr_flagshigh = flags >> 16;
804 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
805 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
813 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
815 struct ifaliasreq ifra;
816 struct sockaddr_in *in;
818 bzero(&ifra, sizeof(ifra));
819 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
822 in = (struct sockaddr_in *)&ifra.ifra_addr;
823 in->sin_family = AF_INET;
824 in->sin_len = sizeof(*in);
825 in->sin_addr.s_addr = addr;
829 in = (struct sockaddr_in *)&ifra.ifra_mask;
830 in->sin_len = sizeof(*in);
831 in->sin_addr.s_addr = mask;
834 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
835 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
843 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
848 bzero(&ifbr, sizeof(ifbr));
849 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
852 bzero(&ifd, sizeof(ifd));
853 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
854 ifd.ifd_cmd = BRDGADD;
855 ifd.ifd_len = sizeof(ifbr);
856 ifd.ifd_data = &ifbr;
858 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
860 * 'errno == EEXIST' means that the tap(4) is already
861 * a member of the bridge(4)
863 if (errno != EEXIST) {
864 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
871 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
873 /* XXX major()/minor() can't be used in vkernel */
874 #define TAPDEV_MAJOR(x) ((int)(((u_int)(x) >> 8) & 0xff))
875 #define TAPDEV_MINOR(x) ((int)((x) & 0xffff00ff))
877 #ifndef TAP_CDEV_MAJOR
878 #define TAP_CDEV_MAJOR 149
882 * Locate the first unused tap(4) device file if auto mode is requested,
883 * or open the user supplied device file, and bring up the corresponding
886 * NOTE: Only tap(4) device file is supported currently
890 netif_open_tap(const char *netif, int *tap_unit, int s)
892 char tap_dev[MAXPATHLEN];
898 if (strcmp(netif, "auto") == 0) {
903 * Find first unused tap(4) device file
906 snprintf(tap_dev, sizeof(tap_dev), "/dev/tap%d", i);
907 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
908 if (tap_fd >= 0 || errno == ENOENT)
913 warnc(lasterr, "Unable to find a free tap(4)");
918 * User supplied tap(4) device file or unix socket.
920 if (netif[0] == '/') /* Absolute path */
921 strlcpy(tap_dev, netif, sizeof(tap_dev));
923 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
925 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
928 * If we cannot open normally try to connect to it.
931 tap_fd = unix_connect(tap_dev);
934 warn("Unable to open %s", tap_dev);
940 * Check whether the device file is a tap(4)
943 if (fstat(tap_fd, &st) == 0 && S_ISCHR(st.st_mode) &&
944 TAPDEV_MAJOR(st.st_rdev) == TAP_CDEV_MAJOR) {
945 *tap_unit = TAPDEV_MINOR(st.st_rdev);
948 * Bring up the corresponding tap(4) interface
950 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
952 } else if (S_ISSOCK(st.st_mode)) {
954 * Special socket connection (typically to vknet). We
955 * do not have to do anything.
959 warnx("%s is not a tap(4) device", tap_dev);
971 unix_connect(const char *path)
973 struct sockaddr_un sunx;
979 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
980 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
981 ++len; /* include nul */
982 sunx.sun_family = AF_UNIX;
985 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
988 if (connect(net_fd, (void *)&sunx, len) < 0) {
992 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
993 if (fstat(net_fd, &st) == 0)
994 printf("Network socket buffer: %d bytes\n", st.st_blksize);
995 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1001 #undef TAPDEV_OFLAGS
1004 * Following syntax is supported,
1005 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1007 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1008 * tap(4)'s netmask len is z
1010 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1011 * pseudo netif's address is y.y.y.y
1013 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1014 * pseudo netif's address is y.y.y.y
1015 * tap(4) and pseudo netif's netmask len are z
1017 * 5) bridgeX tap(4) will be added to bridgeX
1019 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1020 * pseudo netif's address is y.y.y.y
1022 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1023 * pseudo netif's address is y.y.y.y
1024 * pseudo netif's netmask len is z
1028 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1030 in_addr_t tap_addr, netmask, netif_addr;
1031 int next_netif_addr;
1032 char *tok, *masklen_str, *ifbridge;
1037 tok = strtok(NULL, ":/");
1040 * Nothing special, simply use tap(4) as backend
1045 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1047 * tap(4)'s address is supplied
1052 * If there is next token, then it may be pseudo
1053 * netif's address or netmask len for tap(4)
1055 next_netif_addr = 0;
1058 * Not tap(4)'s address, assume it as a bridge(4)
1065 * If there is next token, then it must be pseudo
1068 next_netif_addr = 1;
1071 netmask = netif_addr = 0;
1073 tok = strtok(NULL, ":/");
1077 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1078 if (next_netif_addr) {
1079 warnx("Invalid pseudo netif address: %s", tok);
1085 * Current token is not address, then it must be netmask len
1090 * Current token is pseudo netif address, if there is next token
1091 * it must be netmask len
1093 masklen_str = strtok(NULL, "/");
1096 /* Calculate netmask */
1097 if (masklen_str != NULL) {
1100 masklen = strtoul(masklen_str, NULL, 10);
1101 if (masklen < 32 && masklen > 0) {
1102 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1105 warnx("Invalid netmask len: %lu", masklen);
1110 /* Make sure there is no more token left */
1111 if (strtok(NULL, ":/") != NULL) {
1112 warnx("Invalid argument to '-I'");
1119 } else if (ifbridge == NULL) {
1120 /* Set tap(4) address/netmask */
1121 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1124 /* Tie tap(4) to bridge(4) */
1125 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1135 * NetifInfo[] will be filled for pseudo netif initialization.
1136 * NetifNum will be bumped to reflect the number of valid entries
1141 init_netif(char *netifExp[], int netifExpNum)
1145 if (netifExpNum == 0)
1148 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1152 for (i = 0; i < netifExpNum; ++i) {
1153 struct vknetif_info *info;
1154 in_addr_t netif_addr, netif_mask;
1155 int tap_fd, tap_unit;
1158 netif = strtok(netifExp[i], ":");
1159 if (netif == NULL) {
1160 warnx("Invalid argument to '-I'");
1165 * Open tap(4) device file and bring up the
1166 * corresponding interface
1168 tap_fd = netif_open_tap(netif, &tap_unit, s);
1173 * Initialize tap(4) and get address/netmask
1176 * NB: Rest part of netifExp[i] is passed
1177 * to netif_init_tap() implicitly.
1179 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1181 * NB: Closing tap(4) device file will bring
1182 * down the corresponding interface
1188 info = &NetifInfo[NetifNum];
1189 info->tap_fd = tap_fd;
1190 info->tap_unit = tap_unit;
1191 info->netif_addr = netif_addr;
1192 info->netif_mask = netif_mask;
1195 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1208 if (pid_file != NULL) {
1210 fp = fopen(pid_file, "w");
1213 fprintf(fp, "%ld\n", (long)self);
1217 perror("Warning: couldn't open pidfile");
1226 if (pid_file != NULL) {
1227 if ( unlink(pid_file) != 0 )
1228 perror("Warning: couldn't remove pidfile");
1234 usage(const char *ctl, ...)
1239 vfprintf(stderr, ctl, va);
1241 fprintf(stderr, "\n");
1248 kprintf("cpu reset, rebooting vkernel\n");
1251 execv(save_av[0], save_av);
1257 kprintf("cpu halt, exiting vkernel\n");
1265 switch(lwp_cpu_lock) {
1268 kprintf("Locking CPU%d to real cpu %d\n",
1270 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1272 if (next_cpu >= real_ncpus)
1275 case LCL_SINGLE_CPU:
1277 kprintf("Locking CPU%d to real cpu %d\n",
1279 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1282 /* do not map virtual cpus to real cpus */