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>
82 vm_paddr_t phys_avail[16];
84 vm_paddr_t Maxmem_bytes;
87 struct vkdisk_info DiskInfo[VKDISK_MAX];
89 struct vknetif_info NetifInfo[VKNETIF_MAX];
95 vm_offset_t virtual_start;
96 vm_offset_t virtual_end;
97 vm_offset_t virtual2_start;
98 vm_offset_t virtual2_end;
99 vm_offset_t kernel_vm_end;
100 vm_offset_t crashdumpmap;
101 vm_offset_t clean_sva;
102 vm_offset_t clean_eva;
103 struct msgbuf *msgbufp;
106 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
107 u_int cpu_feature; /* XXX */
109 int64_t tsc_frequency;
110 int optcpus; /* number of cpus - see mp_start() */
111 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
112 int real_ncpus; /* number of real CPUs */
113 int next_cpu; /* next real CPU to lock a virtual CPU to */
115 int via_feature_xcrypt = 0; /* XXX */
116 int via_feature_rng = 0; /* XXX */
118 struct privatespace *CPU_prvspace;
120 static struct trapframe proc0_tf;
121 static void *proc0paddr;
123 static void init_sys_memory(char *imageFile);
124 static void init_kern_memory(void);
125 static void init_globaldata(void);
126 static void init_vkernel(void);
127 static void init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type);
128 static void init_netif(char *netifExp[], int netifFileNum);
129 static void writepid(void);
130 static void cleanpid(void);
131 static int unix_connect(const char *path);
132 static void usage_err(const char *ctl, ...);
133 static void usage_help(_Bool);
136 static char **save_av;
139 * Kernel startup for virtual kernels - standard main()
142 main(int ac, char **av)
144 char *memImageFile = NULL;
145 char *netifFile[VKNETIF_MAX];
146 char *diskFile[VKDISK_MAX];
147 char *cdFile[VKDISK_MAX];
150 int netifFileNum = 0;
153 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
159 int real_vkernel_enable;
169 kernel_mem_readonly = 1;
173 lwp_cpu_lock = LCL_NONE;
175 real_vkernel_enable = 0;
176 vsize = sizeof(real_vkernel_enable);
177 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
179 if (real_vkernel_enable == 0) {
180 errx(1, "vm.vkernel_enable is 0, must be set "
181 "to 1 to execute a vkernel!");
185 vsize = sizeof(real_ncpus);
186 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
191 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:U")) != -1) {
195 * name=value:name=value:name=value...
198 * Allow values to be quoted but note that shells
199 * may remove the quotes, so using this feature
200 * to embed colons may require a backslash.
204 kern_envp = malloc(n + 2);
205 for (i = j = 0; i < n; ++i) {
206 if (optarg[i] == '"')
208 else if (optarg[i] == '\'')
210 else if (isq == 0 && optarg[i] == ':')
213 kern_envp[j++] = optarg[i];
219 boothowto |= RB_SINGLE;
225 memImageFile = optarg;
228 if (netifFileNum < VKNETIF_MAX)
229 netifFile[netifFileNum++] = strdup(optarg);
234 if (diskFileNum + cdFileNum < VKDISK_MAX)
235 diskFile[diskFileNum++] = strdup(optarg);
240 if (diskFileNum + cdFileNum < VKDISK_MAX)
241 cdFile[cdFileNum++] = strdup(optarg);
244 Maxmem_bytes = strtoull(optarg, &suffix, 0);
261 usage_err("Bad maxmem option");
269 if (strncmp("map", optarg, 3) == 0) {
270 lwp_cpu_lock = LCL_PER_CPU;
271 if (optarg[3] == ',') {
272 next_cpu = strtol(optarg+4, &endp, 0);
274 usage_err("Bad target CPU number at '%s'", endp);
278 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
279 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
280 } else if (strncmp("any", optarg, 3) == 0) {
281 lwp_cpu_lock = LCL_NONE;
283 lwp_cpu_lock = LCL_SINGLE_CPU;
284 next_cpu = strtol(optarg, &endp, 0);
286 usage_err("Bad target CPU number at '%s'", endp);
287 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
288 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
293 * This value is set up by mp_start(), don't just
297 optcpus = strtol(optarg, NULL, 0);
298 if (optcpus < 1 || optcpus > MAXCPU)
299 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
301 if (strtol(optarg, NULL, 0) != 1) {
302 usage_err("You built a UP vkernel, only 1 cpu!");
311 kernel_mem_readonly = 0;
323 init_sys_memory(memImageFile);
333 vsize = sizeof(tsc_present);
334 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
335 vsize = sizeof(tsc_frequency);
336 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
338 cpu_feature |= CPUID_TSC;
343 vsize = sizeof(supports_sse);
345 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
346 init_fpu(supports_sse);
348 cpu_feature |= CPUID_SSE | CPUID_FXSR;
351 * We boot from the first installed disk.
353 if (bootOnDisk == 1) {
354 init_disk(diskFile, diskFileNum, VKD_DISK);
355 init_disk(cdFile, cdFileNum, VKD_CD);
357 init_disk(cdFile, cdFileNum, VKD_CD);
358 init_disk(diskFile, diskFileNum, VKD_DISK);
360 init_netif(netifFile, netifFileNum);
368 * Initialize system memory. This is the virtual kernel's 'RAM'.
372 init_sys_memory(char *imageFile)
379 * Figure out the system memory image size. If an image file was
380 * specified and -m was not specified, use the image file's size.
383 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
384 Maxmem_bytes = (vm_paddr_t)st.st_size;
385 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
387 err(1, "Cannot create new memory file %s unless "
388 "system memory size is specified with -m",
394 * Maxmem must be known at this time
396 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
397 err(1, "Bad maxmem specification: 32MB minimum, "
398 "multiples of %dMB only",
399 SEG_SIZE / 1024 / 1024);
404 * Generate an image file name if necessary, then open/create the
405 * file exclusively locked. Do not allow multiple virtual kernels
406 * to use the same image file.
408 if (imageFile == NULL) {
409 for (i = 0; i < 1000000; ++i) {
410 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
412 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
413 if (fd < 0 && errno == EWOULDBLOCK) {
420 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
422 printf("Using memory file: %s\n", imageFile);
423 if (fd < 0 || fstat(fd, &st) < 0) {
424 err(1, "Unable to open/create %s", imageFile);
429 * Truncate or extend the file as necessary. Clean out the contents
430 * of the file, we want it to be full of holes so we don't waste
431 * time reading in data from an old file that we no longer care
435 ftruncate(fd, Maxmem_bytes);
438 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
443 * Initialize kernel memory. This reserves kernel virtual memory by using
449 init_kern_memory(void)
455 char *topofstack = &dummy;
460 * Memory map our kernel virtual memory space. Note that the
461 * kernel image itself is not made part of this memory for the
464 * The memory map must be segment-aligned so we can properly
467 * If the system kernel has a different MAXDSIZ, it might not
468 * be possible to map kernel memory in its prefered location.
469 * Try a number of different locations.
471 try = (void *)0x40000000;
473 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
474 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
475 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
479 if (base != MAP_FAILED)
480 munmap(base, KERNEL_KVA_SIZE);
481 try = (char *)try + 0x10000000;
484 err(1, "Unable to mmap() kernel virtual memory!");
487 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
488 KvaStart = (vm_offset_t)base;
489 KvaSize = KERNEL_KVA_SIZE;
490 KvaEnd = KvaStart + KvaSize;
492 /* cannot use kprintf yet */
493 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
496 * Create a top-level page table self-mapping itself.
498 * Initialize the page directory at physical page index 0 to point
499 * to an array of page table pages starting at physical page index 1
501 lseek(MemImageFd, 0L, 0);
502 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
503 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
504 write(MemImageFd, &pte, sizeof(pte));
508 * Initialize the PTEs in the page table pages required to map the
509 * page table itself. This includes mapping the page directory page
510 * at the base so we go one more loop then normal.
512 lseek(MemImageFd, PAGE_SIZE, 0);
513 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
514 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
515 write(MemImageFd, &pte, sizeof(pte));
519 * Initialize remaining PTEs to 0. We may be reusing a memory image
520 * file. This is approximately a megabyte.
522 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
523 zero = malloc(PAGE_SIZE);
524 bzero(zero, PAGE_SIZE);
526 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
527 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
532 * Enable the page table and calculate pointers to our self-map
533 * for easy kernel page table manipulation.
535 * KernelPTA must be offset so we can do direct VA translations
537 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
538 0 | VPTE_R | VPTE_W | VPTE_V);
539 KernelPTD = (vpte_t *)base; /* pg directory */
540 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
541 KernelPTA -= KvaStart >> PAGE_SHIFT;
544 * phys_avail[] represents unallocated physical memory. MI code
545 * will use phys_avail[] to create the vm_page array.
547 phys_avail[0] = PAGE_SIZE +
548 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
549 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
550 phys_avail[1] = Maxmem_bytes;
553 * (virtual_start, virtual_end) represent unallocated kernel virtual
554 * memory. MI code will create kernel_map using these parameters.
556 virtual_start = KvaStart + PAGE_SIZE +
557 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
558 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
559 virtual_end = KvaStart + KERNEL_KVA_SIZE;
562 * kernel_vm_end could be set to virtual_end but we want some
563 * indication of how much of the kernel_map we've used, so
564 * set it low and let pmap_growkernel increase it even though we
565 * don't need to create any new page table pages.
567 kernel_vm_end = virtual_start;
570 * Allocate space for process 0's UAREA.
572 proc0paddr = (void *)virtual_start;
573 for (i = 0; i < UPAGES; ++i) {
574 pmap_kenter_quick(virtual_start, phys_avail[0]);
575 virtual_start += PAGE_SIZE;
576 phys_avail[0] += PAGE_SIZE;
582 crashdumpmap = virtual_start;
583 virtual_start += MAXDUMPPGS * PAGE_SIZE;
586 * msgbufp maps the system message buffer
588 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
589 msgbufp = (void *)virtual_start;
590 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
591 pmap_kenter_quick(virtual_start, phys_avail[0]);
592 virtual_start += PAGE_SIZE;
593 phys_avail[0] += PAGE_SIZE;
595 msgbufinit(msgbufp, MSGBUF_SIZE);
598 * used by kern_memio for /dev/mem access
600 ptvmmap = (caddr_t)virtual_start;
601 virtual_start += PAGE_SIZE;
604 * Bootstrap the kernel_pmap
610 * Map the per-cpu globaldata for cpu #0. Allocate the space using
611 * virtual_start and phys_avail[0]
615 init_globaldata(void)
622 * Reserve enough KVA to cover possible cpus. This is a considerable
623 * amount of KVA since the privatespace structure includes two
624 * whole page table mappings.
626 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
627 CPU_prvspace = (void *)virtual_start;
628 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
631 * Allocate enough physical memory to cover the mdglobaldata
632 * portion of the space and the idle stack and map the pages
633 * into KVA. For cpu #0 only.
635 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
637 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
638 pmap_kenter_quick(va, pa);
639 phys_avail[0] += PAGE_SIZE;
641 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
643 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
644 pmap_kenter_quick(va, pa);
645 phys_avail[0] += PAGE_SIZE;
649 * Setup the %fs for cpu #0. The mycpu macro works after this
650 * point. Note that %gs is used by pthreads.
652 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
656 * Initialize very low level systems including thread0, proc0, etc.
662 struct mdglobaldata *gd;
664 gd = &CPU_prvspace[0].mdglobaldata;
665 bzero(gd, sizeof(*gd));
667 gd->mi.gd_curthread = &thread0;
668 thread0.td_gd = &gd->mi;
670 ncpus2 = 1; /* rounded down power of 2 */
671 ncpus_fit = 1; /* rounded up power of 2 */
672 /* ncpus2_mask and ncpus_fit_mask are 0 */
674 gd->mi.gd_prvspace = &CPU_prvspace[0];
675 mi_gdinit(&gd->mi, 0);
677 mi_proc0init(&gd->mi, proc0paddr);
678 lwp0.lwp_md.md_regs = &proc0_tf;
683 #if 0 /* #ifdef DDB */
685 if (boothowto & RB_KDB)
686 Debugger("Boot flags requested debugger");
690 initializecpu(); /* Initialize CPU registers */
692 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
696 * Map the message buffer
698 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
699 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
700 msgbufinit(msgbufp, MSGBUF_SIZE);
703 thread0.td_pcb_cr3 ... MMU
704 lwp0.lwp_md.md_regs = &proc0_tf;
709 * Filesystem image paths for the virtual kernel are optional.
710 * If specified they each should point to a disk image,
711 * the first of which will become the root disk.
713 * The virtual kernel caches data from our 'disk' just like a normal kernel,
714 * so we do not really want the real kernel to cache the data too. Use
715 * O_DIRECT to remove the duplication.
719 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
723 if (diskFileNum == 0)
726 for(i=0; i < diskFileNum; i++){
731 warnx("Invalid argument to '-r'");
735 if (DiskNum < VKDISK_MAX) {
737 struct vkdisk_info* info = NULL;
741 if (type == VKD_DISK)
742 fd = open(fname, O_RDWR|O_DIRECT, 0644);
744 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
745 if (fd < 0 || fstat(fd, &st) < 0) {
746 err(1, "Unable to open/create %s", fname);
749 if (S_ISREG(st.st_mode)) {
750 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
751 errx(1, "Disk image %s is already "
757 info = &DiskInfo[DiskNum];
763 memcpy(info->fname, fname, l);
766 if (type == VKD_CD) {
767 rootdevnames[0] = "cd9660:vcd0a";
768 } else if (type == VKD_DISK) {
769 rootdevnames[0] = "ufs:vkd0s0a";
770 rootdevnames[1] = "ufs:vkd0s1a";
776 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
784 netif_set_tapflags(int tap_unit, int f, int s)
789 bzero(&ifr, sizeof(ifr));
791 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
792 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
793 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
800 * If the flags are already set/cleared, then we return
801 * immediately to avoid extra syscalls
803 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
807 if ((flags & f) == 0)
818 * Fix up ifreq.ifr_name, since it may be trashed
819 * in previous ioctl(SIOCGIFFLAGS)
821 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
823 ifr.ifr_flags = flags & 0xffff;
824 ifr.ifr_flagshigh = flags >> 16;
825 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
826 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
834 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
836 struct ifaliasreq ifra;
837 struct sockaddr_in *in;
839 bzero(&ifra, sizeof(ifra));
840 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
843 in = (struct sockaddr_in *)&ifra.ifra_addr;
844 in->sin_family = AF_INET;
845 in->sin_len = sizeof(*in);
846 in->sin_addr.s_addr = addr;
850 in = (struct sockaddr_in *)&ifra.ifra_mask;
851 in->sin_len = sizeof(*in);
852 in->sin_addr.s_addr = mask;
855 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
856 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
864 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
869 bzero(&ifbr, sizeof(ifbr));
870 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
873 bzero(&ifd, sizeof(ifd));
874 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
875 ifd.ifd_cmd = BRDGADD;
876 ifd.ifd_len = sizeof(ifbr);
877 ifd.ifd_data = &ifbr;
879 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
881 * 'errno == EEXIST' means that the tap(4) is already
882 * a member of the bridge(4)
884 if (errno != EEXIST) {
885 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
892 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
895 * Locate the first unused tap(4) device file if auto mode is requested,
896 * or open the user supplied device file, and bring up the corresponding
899 * NOTE: Only tap(4) device file is supported currently
903 netif_open_tap(const char *netif, int *tap_unit, int s)
905 char tap_dev[MAXPATHLEN];
912 if (strcmp(netif, "auto") == 0) {
914 * Find first unused tap(4) device file
916 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
918 warnc(errno, "Unable to find a free tap(4)");
923 * User supplied tap(4) device file or unix socket.
925 if (netif[0] == '/') /* Absolute path */
926 strlcpy(tap_dev, netif, sizeof(tap_dev));
928 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
930 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
933 * If we cannot open normally try to connect to it.
936 tap_fd = unix_connect(tap_dev);
939 warn("Unable to open %s", tap_dev);
945 * Check whether the device file is a tap(4)
947 if (fstat(tap_fd, &st) < 0) {
949 } else if (S_ISCHR(st.st_mode)) {
950 dname = fdevname(tap_fd);
952 dname = strstr(dname, "tap");
955 * Bring up the corresponding tap(4) interface
957 *tap_unit = strtol(dname + 3, NULL, 10);
958 printf("TAP UNIT %d\n", *tap_unit);
959 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
966 } else if (S_ISSOCK(st.st_mode)) {
968 * Special socket connection (typically to vknet). We
969 * do not have to do anything.
977 warnx("%s is not a tap(4) device or socket", tap_dev);
986 unix_connect(const char *path)
988 struct sockaddr_un sunx;
994 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
995 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
996 ++len; /* include nul */
997 sunx.sun_family = AF_UNIX;
1000 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1003 if (connect(net_fd, (void *)&sunx, len) < 0) {
1007 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1008 if (fstat(net_fd, &st) == 0)
1009 printf("Network socket buffer: %d bytes\n", st.st_blksize);
1010 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1016 #undef TAPDEV_OFLAGS
1019 * Following syntax is supported,
1020 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1022 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1023 * tap(4)'s netmask len is z
1025 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1026 * pseudo netif's address is y.y.y.y
1028 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1029 * pseudo netif's address is y.y.y.y
1030 * tap(4) and pseudo netif's netmask len are z
1032 * 5) bridgeX tap(4) will be added to bridgeX
1034 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1035 * pseudo netif's address is y.y.y.y
1037 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1038 * pseudo netif's address is y.y.y.y
1039 * pseudo netif's netmask len is z
1043 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1045 in_addr_t tap_addr, netmask, netif_addr;
1046 int next_netif_addr;
1047 char *tok, *masklen_str, *ifbridge;
1052 tok = strtok(NULL, ":/");
1055 * Nothing special, simply use tap(4) as backend
1060 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1062 * tap(4)'s address is supplied
1067 * If there is next token, then it may be pseudo
1068 * netif's address or netmask len for tap(4)
1070 next_netif_addr = 0;
1073 * Not tap(4)'s address, assume it as a bridge(4)
1080 * If there is next token, then it must be pseudo
1083 next_netif_addr = 1;
1086 netmask = netif_addr = 0;
1088 tok = strtok(NULL, ":/");
1092 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1093 if (next_netif_addr) {
1094 warnx("Invalid pseudo netif address: %s", tok);
1100 * Current token is not address, then it must be netmask len
1105 * Current token is pseudo netif address, if there is next token
1106 * it must be netmask len
1108 masklen_str = strtok(NULL, "/");
1111 /* Calculate netmask */
1112 if (masklen_str != NULL) {
1115 masklen = strtoul(masklen_str, NULL, 10);
1116 if (masklen < 32 && masklen > 0) {
1117 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1120 warnx("Invalid netmask len: %lu", masklen);
1125 /* Make sure there is no more token left */
1126 if (strtok(NULL, ":/") != NULL) {
1127 warnx("Invalid argument to '-I'");
1134 } else if (ifbridge == NULL) {
1135 /* Set tap(4) address/netmask */
1136 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1139 /* Tie tap(4) to bridge(4) */
1140 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1150 * NetifInfo[] will be filled for pseudo netif initialization.
1151 * NetifNum will be bumped to reflect the number of valid entries
1156 init_netif(char *netifExp[], int netifExpNum)
1160 if (netifExpNum == 0)
1163 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1167 for (i = 0; i < netifExpNum; ++i) {
1168 struct vknetif_info *info;
1169 in_addr_t netif_addr, netif_mask;
1170 int tap_fd, tap_unit;
1173 netif = strtok(netifExp[i], ":");
1174 if (netif == NULL) {
1175 warnx("Invalid argument to '-I'");
1180 * Open tap(4) device file and bring up the
1181 * corresponding interface
1183 tap_fd = netif_open_tap(netif, &tap_unit, s);
1188 * Initialize tap(4) and get address/netmask
1191 * NB: Rest part of netifExp[i] is passed
1192 * to netif_init_tap() implicitly.
1194 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1196 * NB: Closing tap(4) device file will bring
1197 * down the corresponding interface
1203 info = &NetifInfo[NetifNum];
1204 info->tap_fd = tap_fd;
1205 info->tap_unit = tap_unit;
1206 info->netif_addr = netif_addr;
1207 info->netif_mask = netif_mask;
1210 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1223 if (pid_file != NULL) {
1225 fp = fopen(pid_file, "w");
1228 fprintf(fp, "%ld\n", (long)self);
1232 perror("Warning: couldn't open pidfile");
1241 if (pid_file != NULL) {
1242 if ( unlink(pid_file) != 0 )
1243 perror("Warning: couldn't remove pidfile");
1249 usage_err(const char *ctl, ...)
1254 vfprintf(stderr, ctl, va);
1256 fprintf(stderr, "\n");
1262 usage_help(_Bool help)
1264 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1265 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1266 "\t[-m size] [-n numcpus] [-p file] [-r file]\n", save_av[0]);
1269 fprintf(stderr, "\nArguments:\n"
1270 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1271 "\t-e\tSpecify an environment to be used by the kernel.\n"
1272 "\t-h\tThis list of options.\n"
1273 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1274 "\t-I\tCreate a virtual network device.\n"
1275 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1276 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1277 "\t-n\tSpecify the number of CPUs you wish to emulate.\n"
1278 "\t-p\tSpecify a file in which to store the process ID.\n"
1279 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1280 "\t-s\tBoot into single-user mode.\n"
1281 "\t-U\tEnable writing to kernel memory and module loading.\n"
1282 "\t-v\tTurn on verbose booting.\n");
1290 kprintf("cpu reset, rebooting vkernel\n");
1293 execv(save_av[0], save_av);
1299 kprintf("cpu halt, exiting vkernel\n");
1307 switch(lwp_cpu_lock) {
1310 kprintf("Locking CPU%d to real cpu %d\n",
1312 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1314 if (next_cpu >= real_ncpus)
1317 case LCL_SINGLE_CPU:
1319 kprintf("Locking CPU%d to real cpu %d\n",
1321 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1324 /* do not map virtual cpus to real cpus */