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
35 #include <sys/types.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
41 #include <sys/random.h>
42 #include <sys/vkernel.h>
44 #include <sys/reboot.h>
46 #include <sys/msgbuf.h>
47 #include <sys/vmspace.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_map.h>
54 #include <sys/mplock2.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 void *dmap_min_address;
108 u_int cpu_feature; /* XXX */
110 int64_t tsc_frequency;
111 int optcpus; /* number of cpus - see mp_start() */
112 int lwp_cpu_lock; /* if/how to lock virtual CPUs to real CPUs */
113 int real_ncpus; /* number of real CPUs */
114 int next_cpu; /* next real CPU to lock a virtual CPU to */
115 int vkernel_b_arg; /* -b argument - no of logical CPU bits - only SMP */
116 int vkernel_B_arg; /* -B argument - no of core bits - only SMP */
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);
134 static void init_locks(void);
137 static char **save_av;
140 * Kernel startup for virtual kernels - standard main()
143 main(int ac, char **av)
145 char *memImageFile = NULL;
146 char *netifFile[VKNETIF_MAX];
147 char *diskFile[VKDISK_MAX];
148 char *cdFile[VKDISK_MAX];
153 int netifFileNum = 0;
156 int bootOnDisk = -1; /* set below to vcd (0) or vkd (1) */
164 int real_vkernel_enable;
179 kernel_mem_readonly = 1;
185 lwp_cpu_lock = LCL_NONE;
187 real_vkernel_enable = 0;
188 vsize = sizeof(real_vkernel_enable);
189 sysctlbyname("vm.vkernel_enable", &real_vkernel_enable, &vsize, NULL,0);
191 if (real_vkernel_enable == 0) {
192 errx(1, "vm.vkernel_enable is 0, must be set "
193 "to 1 to execute a vkernel!");
197 vsize = sizeof(real_ncpus);
198 sysctlbyname("hw.ncpu", &real_ncpus, &vsize, NULL, 0);
203 while ((c = getopt(ac, av, "c:hsvl:m:n:r:e:i:p:I:Ub:B:")) != -1) {
207 * name=value:name=value:name=value...
210 * Allow values to be quoted but note that shells
211 * may remove the quotes, so using this feature
212 * to embed colons may require a backslash.
219 kern_envp = malloc(kenv_size);
220 if (kern_envp == NULL)
221 errx(1, "Couldn't allocate %zd bytes for kern_envp", kenv_size);
223 kenv_size2 = kenv_size + n + 1;
225 if ((tmp = realloc(kern_envp, kenv_size2)) == NULL)
226 errx(1, "Couldn't reallocate %zd bytes for kern_envp", kenv_size2);
228 kenv_size = kenv_size2;
231 for (i = 0, j = pos; i < n; ++i) {
232 if (optarg[i] == '"')
234 else if (optarg[i] == '\'')
236 else if (isq == 0 && optarg[i] == ':')
239 kern_envp[j++] = optarg[i];
246 boothowto |= RB_SINGLE;
252 memImageFile = optarg;
255 if (netifFileNum < VKNETIF_MAX)
256 netifFile[netifFileNum++] = strdup(optarg);
261 if (diskFileNum + cdFileNum < VKDISK_MAX)
262 diskFile[diskFileNum++] = strdup(optarg);
267 if (diskFileNum + cdFileNum < VKDISK_MAX)
268 cdFile[cdFileNum++] = strdup(optarg);
271 Maxmem_bytes = strtoull(optarg, &suffix, 0);
288 usage_err("Bad maxmem option");
296 if (strncmp("map", optarg, 3) == 0) {
297 lwp_cpu_lock = LCL_PER_CPU;
298 if (optarg[3] == ',') {
299 next_cpu = strtol(optarg+4, &endp, 0);
301 usage_err("Bad target CPU number at '%s'", endp);
305 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
306 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
307 } else if (strncmp("any", optarg, 3) == 0) {
308 lwp_cpu_lock = LCL_NONE;
310 lwp_cpu_lock = LCL_SINGLE_CPU;
311 next_cpu = strtol(optarg, &endp, 0);
313 usage_err("Bad target CPU number at '%s'", endp);
314 if (next_cpu < 0 || next_cpu > real_ncpus - 1)
315 usage_err("Bad target CPU, valid range is 0-%d", real_ncpus - 1);
320 * This value is set up by mp_start(), don't just
323 tok = strtok(optarg, ":");
325 optcpus = strtol(tok, NULL, 0);
326 if (optcpus < 1 || optcpus > MAXCPU)
327 usage_err("Bad ncpus, valid range is 1-%d", MAXCPU);
329 /* :lbits argument */
330 tok = strtok(NULL, ":");
332 vkernel_b_arg = strtol(tok, NULL, 0);
334 /* :cbits argument */
335 tok = strtok(NULL, ":");
337 vkernel_B_arg = strtol(tok, NULL, 0);
343 if (strtol(tok, NULL, 0) != 1) {
344 usage_err("You built a UP vkernel, only 1 cpu!");
347 /* :lbits argument */
348 tok = strtok(NULL, ":");
350 usage_err("You built a UP vkernel. No CPU topology available");
352 /* :cbits argument */
353 tok = strtok(NULL, ":");
355 usage_err("You built a UP vkernel. No CPU topology available");
365 kernel_mem_readonly = 0;
377 init_sys_memory(memImageFile);
389 vsize = sizeof(tsc_present);
390 sysctlbyname("hw.tsc_present", &tsc_present, &vsize, NULL, 0);
391 vsize = sizeof(tsc_frequency);
392 sysctlbyname("hw.tsc_frequency", &tsc_frequency, &vsize, NULL, 0);
394 cpu_feature |= CPUID_TSC;
399 vsize = sizeof(supports_sse);
401 sysctlbyname("hw.instruction_sse", &supports_sse, &vsize, NULL, 0);
402 init_fpu(supports_sse);
404 cpu_feature |= CPUID_SSE | CPUID_FXSR;
407 * We boot from the first installed disk.
409 if (bootOnDisk == 1) {
410 init_disk(diskFile, diskFileNum, VKD_DISK);
411 init_disk(cdFile, cdFileNum, VKD_CD);
413 init_disk(cdFile, cdFileNum, VKD_CD);
414 init_disk(diskFile, diskFileNum, VKD_DISK);
416 init_netif(netifFile, netifFileNum);
424 * Initialize system memory. This is the virtual kernel's 'RAM'.
428 init_sys_memory(char *imageFile)
435 * Figure out the system memory image size. If an image file was
436 * specified and -m was not specified, use the image file's size.
438 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
439 Maxmem_bytes = (vm_paddr_t)st.st_size;
440 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
442 errx(1, "Cannot create new memory file %s unless "
443 "system memory size is specified with -m",
449 * Maxmem must be known at this time
451 if (Maxmem_bytes < 64 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
452 errx(1, "Bad maxmem specification: 64MB minimum, "
453 "multiples of %dMB only",
454 SEG_SIZE / 1024 / 1024);
459 * Generate an image file name if necessary, then open/create the
460 * file exclusively locked. Do not allow multiple virtual kernels
461 * to use the same image file.
463 * Don't iterate through a million files if we do not have write
464 * access to the directory, stop if our open() failed on a
465 * non-existant file. Otherwise opens can fail for any number
467 if (imageFile == NULL) {
468 for (i = 0; i < 1000000; ++i) {
469 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
471 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
472 if (fd < 0 && stat(imageFile, &st) == 0) {
479 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
481 fprintf(stderr, "Using memory file: %s\n", imageFile);
482 if (fd < 0 || fstat(fd, &st) < 0) {
483 err(1, "Unable to open/create %s", imageFile);
488 * Truncate or extend the file as necessary. Clean out the contents
489 * of the file, we want it to be full of holes so we don't waste
490 * time reading in data from an old file that we no longer care
494 ftruncate(fd, Maxmem_bytes);
497 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
502 * Initialize kernel memory. This reserves kernel virtual memory by using
508 init_kern_memory(void)
513 char *topofstack = &dummy;
518 * Memory map our kernel virtual memory space. Note that the
519 * kernel image itself is not made part of this memory for the
522 * The memory map must be segment-aligned so we can properly
525 * If the system kernel has a different MAXDSIZ, it might not
526 * be possible to map kernel memory in its prefered location.
527 * Try a number of different locations.
529 try = (void *)(512UL << 30);
531 while ((char *)try + KERNEL_KVA_SIZE < topofstack) {
532 base = mmap(try, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
533 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE,
534 MemImageFd, (off_t)try);
537 if (base != MAP_FAILED)
538 munmap(base, KERNEL_KVA_SIZE);
539 try = (char *)try + (512UL << 30);
542 err(1, "Unable to mmap() kernel virtual memory!");
545 madvise(base, KERNEL_KVA_SIZE, MADV_NOSYNC);
546 KvaStart = (vm_offset_t)base;
547 KvaSize = KERNEL_KVA_SIZE;
548 KvaEnd = KvaStart + KvaSize;
550 /* cannot use kprintf yet */
551 printf("KVM mapped at %p-%p\n", (void *)KvaStart, (void *)KvaEnd);
554 dmap_min_address = mmap(0, DMAP_SIZE, PROT_READ|PROT_WRITE,
555 MAP_NOCORE|MAP_NOSYNC|MAP_SHARED,
557 if (dmap_min_address == MAP_FAILED) {
558 err(1, "Unable to mmap() kernel DMAP region!");
563 pmap_bootstrap((vm_paddr_t *)&firstfree, (int64_t)base);
565 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
566 0 | VPTE_R | VPTE_W | VPTE_V);
569 * phys_avail[] represents unallocated physical memory. MI code
570 * will use phys_avail[] to create the vm_page array.
572 phys_avail[0] = (vm_paddr_t)firstfree;
573 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
574 phys_avail[1] = Maxmem_bytes;
578 * (virtual_start, virtual_end) represent unallocated kernel virtual
579 * memory. MI code will create kernel_map using these parameters.
581 virtual_start = KvaStart + (long)firstfree;
582 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
583 virtual_end = KvaStart + KERNEL_KVA_SIZE;
587 * pmap_growkernel() will set the correct value.
592 * Allocate space for process 0's UAREA.
594 proc0paddr = (void *)virtual_start;
595 for (i = 0; i < UPAGES; ++i) {
596 pmap_kenter_quick(virtual_start, phys_avail[0]);
597 virtual_start += PAGE_SIZE;
598 phys_avail[0] += PAGE_SIZE;
604 crashdumpmap = virtual_start;
605 virtual_start += MAXDUMPPGS * PAGE_SIZE;
608 * msgbufp maps the system message buffer
610 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
611 msgbufp = (void *)virtual_start;
612 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
613 pmap_kenter_quick(virtual_start, phys_avail[0]);
614 virtual_start += PAGE_SIZE;
615 phys_avail[0] += PAGE_SIZE;
617 msgbufinit(msgbufp, MSGBUF_SIZE);
620 * used by kern_memio for /dev/mem access
622 ptvmmap = (caddr_t)virtual_start;
623 virtual_start += PAGE_SIZE;
626 * Bootstrap the kernel_pmap
634 * Map the per-cpu globaldata for cpu #0. Allocate the space using
635 * virtual_start and phys_avail[0]
639 init_globaldata(void)
646 * Reserve enough KVA to cover possible cpus. This is a considerable
647 * amount of KVA since the privatespace structure includes two
648 * whole page table mappings.
650 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
651 CPU_prvspace = (void *)virtual_start;
652 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
655 * Allocate enough physical memory to cover the mdglobaldata
656 * portion of the space and the idle stack and map the pages
657 * into KVA. For cpu #0 only.
659 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
661 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
662 pmap_kenter_quick(va, pa);
663 phys_avail[0] += PAGE_SIZE;
665 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
667 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
668 pmap_kenter_quick(va, pa);
669 phys_avail[0] += PAGE_SIZE;
673 * Setup the %gs for cpu #0. The mycpu macro works after this
674 * point. Note that %fs is used by pthreads.
676 tls_set_gs(&CPU_prvspace[0], sizeof(struct privatespace));
681 * Initialize pool tokens and other necessary locks
689 * Get the initial mplock with a count of 1 for the BSP.
690 * This uses a LOGICAL cpu ID, ie BSP == 0.
692 cpu_get_initial_mplock();
695 /* our token pool needs to work early */
696 lwkt_token_pool_init();
702 * Initialize very low level systems including thread0, proc0, etc.
708 struct mdglobaldata *gd;
710 gd = &CPU_prvspace[0].mdglobaldata;
711 bzero(gd, sizeof(*gd));
713 gd->mi.gd_curthread = &thread0;
714 thread0.td_gd = &gd->mi;
716 ncpus2 = 1; /* rounded down power of 2 */
717 ncpus_fit = 1; /* rounded up power of 2 */
718 /* ncpus2_mask and ncpus_fit_mask are 0 */
720 gd->mi.gd_prvspace = &CPU_prvspace[0];
721 mi_gdinit(&gd->mi, 0);
723 mi_proc0init(&gd->mi, proc0paddr);
724 lwp0.lwp_md.md_regs = &proc0_tf;
729 #if 0 /* #ifdef DDB */
731 if (boothowto & RB_KDB)
732 Debugger("Boot flags requested debugger");
736 initializecpu(); /* Initialize CPU registers */
738 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
742 * Map the message buffer
744 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
745 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
746 msgbufinit(msgbufp, MSGBUF_SIZE);
749 thread0.td_pcb_cr3 ... MMU
750 lwp0.lwp_md.md_regs = &proc0_tf;
755 * Filesystem image paths for the virtual kernel are optional.
756 * If specified they each should point to a disk image,
757 * the first of which will become the root disk.
759 * The virtual kernel caches data from our 'disk' just like a normal kernel,
760 * so we do not really want the real kernel to cache the data too. Use
761 * O_DIRECT to remove the duplication.
765 init_disk(char *diskExp[], int diskFileNum, enum vkdisk_type type)
769 if (diskFileNum == 0)
772 for(i=0; i < diskFileNum; i++){
777 warnx("Invalid argument to '-r'");
781 if (DiskNum < VKDISK_MAX) {
783 struct vkdisk_info* info = NULL;
787 if (type == VKD_DISK)
788 fd = open(fname, O_RDWR|O_DIRECT, 0644);
790 fd = open(fname, O_RDONLY|O_DIRECT, 0644);
791 if (fd < 0 || fstat(fd, &st) < 0) {
792 err(1, "Unable to open/create %s", fname);
795 if (S_ISREG(st.st_mode)) {
796 if (flock(fd, LOCK_EX|LOCK_NB) < 0) {
797 errx(1, "Disk image %s is already "
803 info = &DiskInfo[DiskNum];
809 memcpy(info->fname, fname, l);
812 if (type == VKD_CD) {
813 rootdevnames[0] = "cd9660:vcd0a";
814 } else if (type == VKD_DISK) {
815 rootdevnames[0] = "ufs:vkd0s0a";
816 rootdevnames[1] = "ufs:vkd0s1a";
822 warnx("vkd%d (%s) > VKDISK_MAX", DiskNum, fname);
830 netif_set_tapflags(int tap_unit, int f, int s)
835 bzero(&ifr, sizeof(ifr));
837 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
838 if (ioctl(s, SIOCGIFFLAGS, &ifr) < 0) {
839 warn("tap%d: ioctl(SIOCGIFFLAGS) failed", tap_unit);
846 * If the flags are already set/cleared, then we return
847 * immediately to avoid extra syscalls
849 flags = (ifr.ifr_flags & 0xffff) | (ifr.ifr_flagshigh << 16);
853 if ((flags & f) == 0)
864 * Fix up ifreq.ifr_name, since it may be trashed
865 * in previous ioctl(SIOCGIFFLAGS)
867 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "tap%d", tap_unit);
869 ifr.ifr_flags = flags & 0xffff;
870 ifr.ifr_flagshigh = flags >> 16;
871 if (ioctl(s, SIOCSIFFLAGS, &ifr) < 0) {
872 warn("tap%d: ioctl(SIOCSIFFLAGS) failed", tap_unit);
880 netif_set_tapaddr(int tap_unit, in_addr_t addr, in_addr_t mask, int s)
882 struct ifaliasreq ifra;
883 struct sockaddr_in *in;
885 bzero(&ifra, sizeof(ifra));
886 snprintf(ifra.ifra_name, sizeof(ifra.ifra_name), "tap%d", tap_unit);
889 in = (struct sockaddr_in *)&ifra.ifra_addr;
890 in->sin_family = AF_INET;
891 in->sin_len = sizeof(*in);
892 in->sin_addr.s_addr = addr;
896 in = (struct sockaddr_in *)&ifra.ifra_mask;
897 in->sin_len = sizeof(*in);
898 in->sin_addr.s_addr = mask;
901 if (ioctl(s, SIOCAIFADDR, &ifra) < 0) {
902 warn("tap%d: ioctl(SIOCAIFADDR) failed", tap_unit);
910 netif_add_tap2brg(int tap_unit, const char *ifbridge, int s)
915 bzero(&ifbr, sizeof(ifbr));
916 snprintf(ifbr.ifbr_ifsname, sizeof(ifbr.ifbr_ifsname),
919 bzero(&ifd, sizeof(ifd));
920 strlcpy(ifd.ifd_name, ifbridge, sizeof(ifd.ifd_name));
921 ifd.ifd_cmd = BRDGADD;
922 ifd.ifd_len = sizeof(ifbr);
923 ifd.ifd_data = &ifbr;
925 if (ioctl(s, SIOCSDRVSPEC, &ifd) < 0) {
927 * 'errno == EEXIST' means that the tap(4) is already
928 * a member of the bridge(4)
930 if (errno != EEXIST) {
931 warn("ioctl(%s, SIOCSDRVSPEC) failed", ifbridge);
938 #define TAPDEV_OFLAGS (O_RDWR | O_NONBLOCK)
941 * Locate the first unused tap(4) device file if auto mode is requested,
942 * or open the user supplied device file, and bring up the corresponding
945 * NOTE: Only tap(4) device file is supported currently
949 netif_open_tap(const char *netif, int *tap_unit, int s)
951 char tap_dev[MAXPATHLEN];
958 if (strcmp(netif, "auto") == 0) {
960 * Find first unused tap(4) device file
962 tap_fd = open("/dev/tap", TAPDEV_OFLAGS);
964 warnc(errno, "Unable to find a free tap(4)");
969 * User supplied tap(4) device file or unix socket.
971 if (netif[0] == '/') /* Absolute path */
972 strlcpy(tap_dev, netif, sizeof(tap_dev));
974 snprintf(tap_dev, sizeof(tap_dev), "/dev/%s", netif);
976 tap_fd = open(tap_dev, TAPDEV_OFLAGS);
979 * If we cannot open normally try to connect to it.
982 tap_fd = unix_connect(tap_dev);
985 warn("Unable to open %s", tap_dev);
991 * Check whether the device file is a tap(4)
993 if (fstat(tap_fd, &st) < 0) {
995 } else if (S_ISCHR(st.st_mode)) {
996 dname = fdevname(tap_fd);
998 dname = strstr(dname, "tap");
1001 * Bring up the corresponding tap(4) interface
1003 *tap_unit = strtol(dname + 3, NULL, 10);
1004 printf("TAP UNIT %d\n", *tap_unit);
1005 if (netif_set_tapflags(*tap_unit, IFF_UP, s) == 0)
1012 } else if (S_ISSOCK(st.st_mode)) {
1014 * Special socket connection (typically to vknet). We
1015 * do not have to do anything.
1023 warnx("%s is not a tap(4) device or socket", tap_dev);
1032 unix_connect(const char *path)
1034 struct sockaddr_un sunx;
1037 int sndbuf = 262144;
1040 snprintf(sunx.sun_path, sizeof(sunx.sun_path), "%s", path);
1041 len = offsetof(struct sockaddr_un, sun_path[strlen(sunx.sun_path)]);
1042 ++len; /* include nul */
1043 sunx.sun_family = AF_UNIX;
1046 net_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1049 if (connect(net_fd, (void *)&sunx, len) < 0) {
1053 setsockopt(net_fd, SOL_SOCKET, SO_SNDBUF, &sndbuf, sizeof(sndbuf));
1054 if (fstat(net_fd, &st) == 0)
1055 printf("Network socket buffer: %d bytes\n", st.st_blksize);
1056 fcntl(net_fd, F_SETFL, O_NONBLOCK);
1062 #undef TAPDEV_OFLAGS
1065 * Following syntax is supported,
1066 * 1) x.x.x.x tap(4)'s address is x.x.x.x
1068 * 2) x.x.x.x/z tap(4)'s address is x.x.x.x
1069 * tap(4)'s netmask len is z
1071 * 3) x.x.x.x:y.y.y.y tap(4)'s address is x.x.x.x
1072 * pseudo netif's address is y.y.y.y
1074 * 4) x.x.x.x:y.y.y.y/z tap(4)'s address is x.x.x.x
1075 * pseudo netif's address is y.y.y.y
1076 * tap(4) and pseudo netif's netmask len are z
1078 * 5) bridgeX tap(4) will be added to bridgeX
1080 * 6) bridgeX:y.y.y.y tap(4) will be added to bridgeX
1081 * pseudo netif's address is y.y.y.y
1083 * 7) bridgeX:y.y.y.y/z tap(4) will be added to bridgeX
1084 * pseudo netif's address is y.y.y.y
1085 * pseudo netif's netmask len is z
1089 netif_init_tap(int tap_unit, in_addr_t *addr, in_addr_t *mask, int s)
1091 in_addr_t tap_addr, netmask, netif_addr;
1092 int next_netif_addr;
1093 char *tok, *masklen_str, *ifbridge;
1098 tok = strtok(NULL, ":/");
1101 * Nothing special, simply use tap(4) as backend
1106 if (inet_pton(AF_INET, tok, &tap_addr) > 0) {
1108 * tap(4)'s address is supplied
1113 * If there is next token, then it may be pseudo
1114 * netif's address or netmask len for tap(4)
1116 next_netif_addr = 0;
1119 * Not tap(4)'s address, assume it as a bridge(4)
1126 * If there is next token, then it must be pseudo
1129 next_netif_addr = 1;
1132 netmask = netif_addr = 0;
1134 tok = strtok(NULL, ":/");
1138 if (inet_pton(AF_INET, tok, &netif_addr) <= 0) {
1139 if (next_netif_addr) {
1140 warnx("Invalid pseudo netif address: %s", tok);
1146 * Current token is not address, then it must be netmask len
1151 * Current token is pseudo netif address, if there is next token
1152 * it must be netmask len
1154 masklen_str = strtok(NULL, "/");
1157 /* Calculate netmask */
1158 if (masklen_str != NULL) {
1161 masklen = strtoul(masklen_str, NULL, 10);
1162 if (masklen < 32 && masklen > 0) {
1163 netmask = htonl(~((1LL << (32 - masklen)) - 1)
1166 warnx("Invalid netmask len: %lu", masklen);
1171 /* Make sure there is no more token left */
1172 if (strtok(NULL, ":/") != NULL) {
1173 warnx("Invalid argument to '-I'");
1180 } else if (ifbridge == NULL) {
1181 /* Set tap(4) address/netmask */
1182 if (netif_set_tapaddr(tap_unit, tap_addr, netmask, s) < 0)
1185 /* Tie tap(4) to bridge(4) */
1186 if (netif_add_tap2brg(tap_unit, ifbridge, s) < 0)
1196 * NetifInfo[] will be filled for pseudo netif initialization.
1197 * NetifNum will be bumped to reflect the number of valid entries
1202 init_netif(char *netifExp[], int netifExpNum)
1206 if (netifExpNum == 0)
1209 s = socket(AF_INET, SOCK_DGRAM, 0); /* for ioctl(SIOC) */
1213 for (i = 0; i < netifExpNum; ++i) {
1214 struct vknetif_info *info;
1215 in_addr_t netif_addr, netif_mask;
1216 int tap_fd, tap_unit;
1219 netif = strtok(netifExp[i], ":");
1220 if (netif == NULL) {
1221 warnx("Invalid argument to '-I'");
1226 * Open tap(4) device file and bring up the
1227 * corresponding interface
1229 tap_fd = netif_open_tap(netif, &tap_unit, s);
1234 * Initialize tap(4) and get address/netmask
1237 * NB: Rest part of netifExp[i] is passed
1238 * to netif_init_tap() implicitly.
1240 if (netif_init_tap(tap_unit, &netif_addr, &netif_mask, s) < 0) {
1242 * NB: Closing tap(4) device file will bring
1243 * down the corresponding interface
1249 info = &NetifInfo[NetifNum];
1250 info->tap_fd = tap_fd;
1251 info->tap_unit = tap_unit;
1252 info->netif_addr = netif_addr;
1253 info->netif_mask = netif_mask;
1256 if (NetifNum >= VKNETIF_MAX) /* XXX will this happen? */
1263 * Create the pid file and leave it open and locked while the vkernel is
1264 * running. This allows a script to use /usr/bin/lockf to probe whether
1265 * a vkernel is still running (so as not to accidently kill an unrelated
1266 * process from a stale pid file).
1275 if (pid_file != NULL) {
1276 snprintf(buf, sizeof(buf), "%ld\n", (long)getpid());
1277 fd = open(pid_file, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0666);
1279 if (errno == EWOULDBLOCK) {
1280 perror("Failed to lock pidfile, "
1281 "vkernel already running");
1283 perror("Failed to create pidfile");
1288 write(fd, buf, strlen(buf));
1289 /* leave the file open to maintain the lock */
1297 if (pid_file != NULL) {
1298 if (unlink(pid_file) < 0)
1299 perror("Warning: couldn't remove pidfile");
1305 usage_err(const char *ctl, ...)
1310 vfprintf(stderr, ctl, va);
1312 fprintf(stderr, "\n");
1318 usage_help(_Bool help)
1320 fprintf(stderr, "Usage: %s [-hsUv] [-c file] [-e name=value:name=value:...]\n"
1321 "\t[-i file] [-I interface[:address1[:address2][/netmask]]] [-l cpulock]\n"
1322 "\t[-m size] [-n numcpus[:lbits[:cbits]]]\n"
1323 "\t[-p file] [-r file]\n", save_av[0]);
1326 fprintf(stderr, "\nArguments:\n"
1327 "\t-c\tSpecify a readonly CD-ROM image file to be used by the kernel.\n"
1328 "\t-e\tSpecify an environment to be used by the kernel.\n"
1329 "\t-h\tThis list of options.\n"
1330 "\t-i\tSpecify a memory image file to be used by the virtual kernel.\n"
1331 "\t-I\tCreate a virtual network device.\n"
1332 "\t-l\tSpecify which, if any, real CPUs to lock virtual CPUs to.\n"
1333 "\t-m\tSpecify the amount of memory to be used by the kernel in bytes.\n"
1334 "\t-n\tSpecify the number of CPUs and the topology you wish to emulate:\n"
1335 "\t \t- numcpus - number of cpus\n"
1336 "\t \t- :lbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1337 "\t \t the logical ID. Controls the number of threads/core (0bits - 1 thread, 1bit - 2 threads).\n"
1338 "\t \t- :cbits - specify the number of bits within APICID(=CPUID) needed for representing\n"
1339 "\t \t the core ID. Controls the number of core/package (0bits - 1 core, 1bit - 2 cores).\n"
1340 "\t-p\tSpecify a file in which to store the process ID.\n"
1341 "\t-r\tSpecify a R/W disk image file to be used by the kernel.\n"
1342 "\t-s\tBoot into single-user mode.\n"
1343 "\t-U\tEnable writing to kernel memory and module loading.\n"
1344 "\t-v\tTurn on verbose booting.\n");
1352 kprintf("cpu reset, rebooting vkernel\n");
1355 execv(save_av[0], save_av);
1361 kprintf("cpu halt, exiting vkernel\n");
1369 switch(lwp_cpu_lock) {
1372 kprintf("Locking CPU%d to real cpu %d\n",
1374 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1376 if (next_cpu >= real_ncpus)
1379 case LCL_SINGLE_CPU:
1381 kprintf("Locking CPU%d to real cpu %d\n",
1383 usched_set(getpid(), USCHED_SET_CPU, &next_cpu, sizeof(next_cpu));
1386 /* do not map virtual cpus to real cpus */
1392 * Allocate and free memory for module loading. The loaded module
1393 * has to be placed somewhere near the current kernel binary load
1394 * point or the relocations will not work.
1396 * I'm not sure why this isn't working.
1399 vkernel_module_memory_alloc(vm_offset_t *basep, size_t bytes)
1401 kprintf("module loading for vkernel64's not currently supported\n");
1407 xtra = (PAGE_SIZE - (vm_offset_t)sbrk(0)) & PAGE_MASK;
1408 *basep = (vm_offset_t)sbrk(xtra + bytes) + xtra;
1409 bzero((void *)*basep, bytes);
1411 *basep = (vm_offset_t)mmap((void *)0x000000000, bytes,
1412 PROT_READ|PROT_WRITE|PROT_EXEC,
1413 MAP_ANON|MAP_SHARED, -1, 0);
1414 if ((void *)*basep == MAP_FAILED)
1417 kprintf("basep %p %p %zd\n",
1418 (void *)vkernel_module_memory_alloc, (void *)*basep, bytes);
1424 vkernel_module_memory_free(vm_offset_t base, size_t bytes)
1428 munmap((void *)base, bytes);