sys/amd64/vmm/vmm_dev.c

   1 /*-
   2  * Copyright (c) 2011 NetApp, Inc.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  *
  14  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
  15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  17  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
  18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  24  * SUCH DAMAGE.
  25  *
  26  * $FreeBSD$
  27  */
  28
  29 #include <sys/cdefs.h>
  30 __FBSDID("$FreeBSD$");
  31
  32 #include <sys/param.h>
  33 #include <sys/kernel.h>
  34 #include <sys/queue.h>
  35 #include <sys/lock.h>
  36 #include <sys/mutex.h>
  37 #include <sys/malloc.h>
  38 #include <sys/conf.h>
  39 #include <sys/sysctl.h>
  40 #include <sys/libkern.h>
  41 #include <sys/ioccom.h>
  42 #include <sys/mman.h>
  43 #include <sys/uio.h>
  44
  45 #include <vm/vm.h>
  46 #include <vm/pmap.h>
  47 #include <vm/vm_map.h>
  48
  49 #include <machine/vmparam.h>
  50 #include <machine/vmm.h>
  51 #include <machine/vmm_instruction_emul.h>
  52 #include <machine/vmm_dev.h>
  53
  54 #include "vmm_lapic.h"
  55 #include "vmm_stat.h"
  56 #include "vmm_mem.h"
  57 #include "io/ppt.h"
  58 #include "io/vatpic.h"
  59 #include "io/vioapic.h"
  60 #include "io/vhpet.h"
  61
  62 struct vmmdev_softc {
  63         struct vm       *vm;            /* vm instance cookie */
  64         struct cdev     *cdev;
  65         SLIST_ENTRY(vmmdev_softc) link;
  66         int             flags;
  67 };
  68 #define VSC_LINKED              0x01
  69
  70 static SLIST_HEAD(, vmmdev_softc) head;
  71
  72 static struct mtx vmmdev_mtx;
  73
  74 static MALLOC_DEFINE(M_VMMDEV, "vmmdev", "vmmdev");
  75
  76 SYSCTL_DECL(_hw_vmm);
  77
  78 static struct vmmdev_softc *
  79 vmmdev_lookup(const char *name)
  80 {
  81         struct vmmdev_softc *sc;
  82
  83 #ifdef notyet   /* XXX kernel is not compiled with invariants */
  84         mtx_assert(&vmmdev_mtx, MA_OWNED);
  85 #endif
  86
  87         SLIST_FOREACH(sc, &head, link) {
  88                 if (strcmp(name, vm_name(sc->vm)) == 0)
  89                         break;
  90         }
  91
  92         return (sc);
  93 }
  94
  95 static struct vmmdev_softc *
  96 vmmdev_lookup2(struct cdev *cdev)
  97 {
  98
  99         return (cdev->si_drv1);
 100 }
 101
 102 static int
 103 vmmdev_rw(struct cdev *cdev, struct uio *uio, int flags)
 104 {
 105         int error, off, c, prot;
 106         vm_paddr_t gpa;
 107         void *hpa, *cookie;
 108         struct vmmdev_softc *sc;
 109
 110         static char zerobuf[PAGE_SIZE];
 111
 112         error = 0;
 113         sc = vmmdev_lookup2(cdev);
 114         if (sc == NULL)
 115                 error = ENXIO;
 116
 117         prot = (uio->uio_rw == UIO_WRITE ? VM_PROT_WRITE : VM_PROT_READ);
 118         while (uio->uio_resid > 0 && error == 0) {
 119                 gpa = uio->uio_offset;
 120                 off = gpa & PAGE_MASK;
 121                 c = min(uio->uio_resid, PAGE_SIZE - off);
 122
 123                 /*
 124                  * The VM has a hole in its physical memory map. If we want to
 125                  * use 'dd' to inspect memory beyond the hole we need to
 126                  * provide bogus data for memory that lies in the hole.
 127                  *
 128                  * Since this device does not support lseek(2), dd(1) will
 129                  * read(2) blocks of data to simulate the lseek(2).
 130                  */
 131                 hpa = vm_gpa_hold(sc->vm, gpa, c, prot, &cookie);
 132                 if (hpa == NULL) {
 133                         if (uio->uio_rw == UIO_READ)
 134                                 error = uiomove(zerobuf, c, uio);
 135                         else
 136                                 error = EFAULT;
 137                 } else {
 138                         error = uiomove(hpa, c, uio);
 139                         vm_gpa_release(cookie);
 140                 }
 141         }
 142         return (error);
 143 }
 144
 145 static int
 146 vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
 147              struct thread *td)
 148 {
 149         int error, vcpu, state_changed, size;
 150         cpuset_t *cpuset;
 151         struct vmmdev_softc *sc;
 152         struct vm_memory_segment *seg;
 153         struct vm_register *vmreg;
 154         struct vm_seg_desc *vmsegdesc;
 155         struct vm_run *vmrun;
 156         struct vm_exception *vmexc;
 157         struct vm_lapic_irq *vmirq;
 158         struct vm_lapic_msi *vmmsi;
 159         struct vm_ioapic_irq *ioapic_irq;
 160         struct vm_isa_irq *isa_irq;
 161         struct vm_isa_irq_trigger *isa_irq_trigger;
 162         struct vm_capability *vmcap;
 163         struct vm_pptdev *pptdev;
 164         struct vm_pptdev_mmio *pptmmio;
 165         struct vm_pptdev_msi *pptmsi;
 166         struct vm_pptdev_msix *pptmsix;
 167         struct vm_nmi *vmnmi;
 168         struct vm_stats *vmstats;
 169         struct vm_stat_desc *statdesc;
 170         struct vm_x2apic *x2apic;
 171         struct vm_gpa_pte *gpapte;
 172         struct vm_suspend *vmsuspend;
 173         struct vm_gla2gpa *gg;
 174         struct vm_activate_cpu *vac;
 175         struct vm_cpuset *vm_cpuset;
 176
 177         sc = vmmdev_lookup2(cdev);
 178         if (sc == NULL)
 179                 return (ENXIO);
 180
 181         error = 0;
 182         vcpu = -1;
 183         state_changed = 0;
 184
 185         /*
 186          * Some VMM ioctls can operate only on vcpus that are not running.
 187          */
 188         switch (cmd) {
 189         case VM_RUN:
 190         case VM_GET_REGISTER:
 191         case VM_SET_REGISTER:
 192         case VM_GET_SEGMENT_DESCRIPTOR:
 193         case VM_SET_SEGMENT_DESCRIPTOR:
 194         case VM_INJECT_EXCEPTION:
 195         case VM_GET_CAPABILITY:
 196         case VM_SET_CAPABILITY:
 197         case VM_PPTDEV_MSI:
 198         case VM_PPTDEV_MSIX:
 199         case VM_SET_X2APIC_STATE:
 200         case VM_GLA2GPA:
 201         case VM_ACTIVATE_CPU:
 202                 /*
 203                  * XXX fragile, handle with care
 204                  * Assumes that the first field of the ioctl data is the vcpu.
 205                  */
 206                 vcpu = *(int *)data;
 207                 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
 208                         error = EINVAL;
 209                         goto done;
 210                 }
 211
 212                 error = vcpu_set_state(sc->vm, vcpu, VCPU_FROZEN, true);
 213                 if (error)
 214                         goto done;
 215
 216                 state_changed = 1;
 217                 break;
 218
 219         case VM_MAP_PPTDEV_MMIO:
 220         case VM_BIND_PPTDEV:
 221         case VM_UNBIND_PPTDEV:
 222         case VM_MAP_MEMORY:
 223         case VM_REINIT:
 224                 /*
 225                  * ioctls that operate on the entire virtual machine must
 226                  * prevent all vcpus from running.
 227                  */
 228                 error = 0;
 229                 for (vcpu = 0; vcpu < VM_MAXCPU; vcpu++) {
 230                         error = vcpu_set_state(sc->vm, vcpu, VCPU_FROZEN, true);
 231                         if (error)
 232                                 break;
 233                 }
 234
 235                 if (error) {
 236                         while (--vcpu >= 0)
 237                                 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
 238                         goto done;
 239                 }
 240
 241                 state_changed = 2;
 242                 break;
 243
 244         default:
 245                 break;
 246         }
 247
 248         switch(cmd) {
 249         case VM_RUN:
 250                 vmrun = (struct vm_run *)data;
 251                 error = vm_run(sc->vm, vmrun);
 252                 break;
 253         case VM_SUSPEND:
 254                 vmsuspend = (struct vm_suspend *)data;
 255                 error = vm_suspend(sc->vm, vmsuspend->how);
 256                 break;
 257         case VM_REINIT:
 258                 error = vm_reinit(sc->vm);
 259                 break;
 260         case VM_STAT_DESC: {
 261                 statdesc = (struct vm_stat_desc *)data;
 262                 error = vmm_stat_desc_copy(statdesc->index,
 263                                         statdesc->desc, sizeof(statdesc->desc));
 264                 break;
 265         }
 266         case VM_STATS: {
 267                 CTASSERT(MAX_VM_STATS >= MAX_VMM_STAT_ELEMS);
 268                 vmstats = (struct vm_stats *)data;
 269                 getmicrotime(&vmstats->tv);
 270                 error = vmm_stat_copy(sc->vm, vmstats->cpuid,
 271                                       &vmstats->num_entries, vmstats->statbuf);
 272                 break;
 273         }
 274         case VM_PPTDEV_MSI:
 275                 pptmsi = (struct vm_pptdev_msi *)data;
 276                 error = ppt_setup_msi(sc->vm, pptmsi->vcpu,
 277                                       pptmsi->bus, pptmsi->slot, pptmsi->func,
 278                                       pptmsi->addr, pptmsi->msg,
 279                                       pptmsi->numvec);
 280                 break;
 281         case VM_PPTDEV_MSIX:
 282                 pptmsix = (struct vm_pptdev_msix *)data;
 283                 error = ppt_setup_msix(sc->vm, pptmsix->vcpu,
 284                                        pptmsix->bus, pptmsix->slot,
 285                                        pptmsix->func, pptmsix->idx,
 286                                        pptmsix->addr, pptmsix->msg,
 287                                        pptmsix->vector_control);
 288                 break;
 289         case VM_MAP_PPTDEV_MMIO:
 290                 pptmmio = (struct vm_pptdev_mmio *)data;
 291                 error = ppt_map_mmio(sc->vm, pptmmio->bus, pptmmio->slot,
 292                                      pptmmio->func, pptmmio->gpa, pptmmio->len,
 293                                      pptmmio->hpa);
 294                 break;
 295         case VM_BIND_PPTDEV:
 296                 pptdev = (struct vm_pptdev *)data;
 297                 error = vm_assign_pptdev(sc->vm, pptdev->bus, pptdev->slot,
 298                                          pptdev->func);
 299                 break;
 300         case VM_UNBIND_PPTDEV:
 301                 pptdev = (struct vm_pptdev *)data;
 302                 error = vm_unassign_pptdev(sc->vm, pptdev->bus, pptdev->slot,
 303                                            pptdev->func);
 304                 break;
 305         case VM_INJECT_EXCEPTION:
 306                 vmexc = (struct vm_exception *)data;
 307                 error = vm_inject_exception(sc->vm, vmexc->cpuid, vmexc);
 308                 break;
 309         case VM_INJECT_NMI:
 310                 vmnmi = (struct vm_nmi *)data;
 311                 error = vm_inject_nmi(sc->vm, vmnmi->cpuid);
 312                 break;
 313         case VM_LAPIC_IRQ:
 314                 vmirq = (struct vm_lapic_irq *)data;
 315                 error = lapic_intr_edge(sc->vm, vmirq->cpuid, vmirq->vector);
 316                 break;
 317         case VM_LAPIC_LOCAL_IRQ:
 318                 vmirq = (struct vm_lapic_irq *)data;
 319                 error = lapic_set_local_intr(sc->vm, vmirq->cpuid,
 320                     vmirq->vector);
 321                 break;
 322         case VM_LAPIC_MSI:
 323                 vmmsi = (struct vm_lapic_msi *)data;
 324                 error = lapic_intr_msi(sc->vm, vmmsi->addr, vmmsi->msg);
 325                 break;
 326         case VM_IOAPIC_ASSERT_IRQ:
 327                 ioapic_irq = (struct vm_ioapic_irq *)data;
 328                 error = vioapic_assert_irq(sc->vm, ioapic_irq->irq);
 329                 break;
 330         case VM_IOAPIC_DEASSERT_IRQ:
 331                 ioapic_irq = (struct vm_ioapic_irq *)data;
 332                 error = vioapic_deassert_irq(sc->vm, ioapic_irq->irq);
 333                 break;
 334         case VM_IOAPIC_PULSE_IRQ:
 335                 ioapic_irq = (struct vm_ioapic_irq *)data;
 336                 error = vioapic_pulse_irq(sc->vm, ioapic_irq->irq);
 337                 break;
 338         case VM_IOAPIC_PINCOUNT:
 339                 *(int *)data = vioapic_pincount(sc->vm);
 340                 break;
 341         case VM_ISA_ASSERT_IRQ:
 342                 isa_irq = (struct vm_isa_irq *)data;
 343                 error = vatpic_assert_irq(sc->vm, isa_irq->atpic_irq);
 344                 if (error == 0 && isa_irq->ioapic_irq != -1)
 345                         error = vioapic_assert_irq(sc->vm,
 346                             isa_irq->ioapic_irq);
 347                 break;
 348         case VM_ISA_DEASSERT_IRQ:
 349                 isa_irq = (struct vm_isa_irq *)data;
 350                 error = vatpic_deassert_irq(sc->vm, isa_irq->atpic_irq);
 351                 if (error == 0 && isa_irq->ioapic_irq != -1)
 352                         error = vioapic_deassert_irq(sc->vm,
 353                             isa_irq->ioapic_irq);
 354                 break;
 355         case VM_ISA_PULSE_IRQ:
 356                 isa_irq = (struct vm_isa_irq *)data;
 357                 error = vatpic_pulse_irq(sc->vm, isa_irq->atpic_irq);
 358                 if (error == 0 && isa_irq->ioapic_irq != -1)
 359                         error = vioapic_pulse_irq(sc->vm, isa_irq->ioapic_irq);
 360                 break;
 361         case VM_ISA_SET_IRQ_TRIGGER:
 362                 isa_irq_trigger = (struct vm_isa_irq_trigger *)data;
 363                 error = vatpic_set_irq_trigger(sc->vm,
 364                     isa_irq_trigger->atpic_irq, isa_irq_trigger->trigger);
 365                 break;
 366         case VM_MAP_MEMORY:
 367                 seg = (struct vm_memory_segment *)data;
 368                 error = vm_malloc(sc->vm, seg->gpa, seg->len);
 369                 break;
 370         case VM_GET_MEMORY_SEG:
 371                 seg = (struct vm_memory_segment *)data;
 372                 seg->len = 0;
 373                 (void)vm_gpabase2memseg(sc->vm, seg->gpa, seg);
 374                 error = 0;
 375                 break;
 376         case VM_GET_REGISTER:
 377                 vmreg = (struct vm_register *)data;
 378                 error = vm_get_register(sc->vm, vmreg->cpuid, vmreg->regnum,
 379                                         &vmreg->regval);
 380                 break;
 381         case VM_SET_REGISTER:
 382                 vmreg = (struct vm_register *)data;
 383                 error = vm_set_register(sc->vm, vmreg->cpuid, vmreg->regnum,
 384                                         vmreg->regval);
 385                 break;
 386         case VM_SET_SEGMENT_DESCRIPTOR:
 387                 vmsegdesc = (struct vm_seg_desc *)data;
 388                 error = vm_set_seg_desc(sc->vm, vmsegdesc->cpuid,
 389                                         vmsegdesc->regnum,
 390                                         &vmsegdesc->desc);
 391                 break;
 392         case VM_GET_SEGMENT_DESCRIPTOR:
 393                 vmsegdesc = (struct vm_seg_desc *)data;
 394                 error = vm_get_seg_desc(sc->vm, vmsegdesc->cpuid,
 395                                         vmsegdesc->regnum,
 396                                         &vmsegdesc->desc);
 397                 break;
 398         case VM_GET_CAPABILITY:
 399                 vmcap = (struct vm_capability *)data;
 400                 error = vm_get_capability(sc->vm, vmcap->cpuid,
 401                                           vmcap->captype,
 402                                           &vmcap->capval);
 403                 break;
 404         case VM_SET_CAPABILITY:
 405                 vmcap = (struct vm_capability *)data;
 406                 error = vm_set_capability(sc->vm, vmcap->cpuid,
 407                                           vmcap->captype,
 408                                           vmcap->capval);
 409                 break;
 410         case VM_SET_X2APIC_STATE:
 411                 x2apic = (struct vm_x2apic *)data;
 412                 error = vm_set_x2apic_state(sc->vm,
 413                                             x2apic->cpuid, x2apic->state);
 414                 break;
 415         case VM_GET_X2APIC_STATE:
 416                 x2apic = (struct vm_x2apic *)data;
 417                 error = vm_get_x2apic_state(sc->vm,
 418                                             x2apic->cpuid, &x2apic->state);
 419                 break;
 420         case VM_GET_GPA_PMAP:
 421                 gpapte = (struct vm_gpa_pte *)data;
 422                 pmap_get_mapping(vmspace_pmap(vm_get_vmspace(sc->vm)),
 423                                  gpapte->gpa, gpapte->pte, &gpapte->ptenum);
 424                 error = 0;
 425                 break;
 426         case VM_GET_HPET_CAPABILITIES:
 427                 error = vhpet_getcap((struct vm_hpet_cap *)data);
 428                 break;
 429         case VM_GLA2GPA: {
 430                 CTASSERT(PROT_READ == VM_PROT_READ);
 431                 CTASSERT(PROT_WRITE == VM_PROT_WRITE);
 432                 CTASSERT(PROT_EXEC == VM_PROT_EXECUTE);
 433                 gg = (struct vm_gla2gpa *)data;
 434                 error = vmm_gla2gpa(sc->vm, gg->vcpuid, &gg->paging, gg->gla,
 435                     gg->prot, &gg->gpa);
 436                 KASSERT(error == 0 || error == 1 || error == -1,
 437                     ("%s: vmm_gla2gpa unknown error %d", __func__, error));
 438                 if (error >= 0) {
 439                         /*
 440                          * error = 0: the translation was successful
 441                          * error = 1: a fault was injected into the guest
 442                          */
 443                         gg->fault = error;
 444                         error = 0;
 445                 } else {
 446                         error = EFAULT;
 447                 }
 448                 break;
 449         }
 450         case VM_ACTIVATE_CPU:
 451                 vac = (struct vm_activate_cpu *)data;
 452                 error = vm_activate_cpu(sc->vm, vac->vcpuid);
 453                 break;
 454         case VM_GET_CPUS:
 455                 error = 0;
 456                 vm_cpuset = (struct vm_cpuset *)data;
 457                 size = vm_cpuset->cpusetsize;
 458                 if (size < sizeof(cpuset_t) || size > CPU_MAXSIZE / NBBY) {
 459                         error = ERANGE;
 460                         break;
 461                 }
 462                 cpuset = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
 463                 if (vm_cpuset->which == VM_ACTIVE_CPUS)
 464                         *cpuset = vm_active_cpus(sc->vm);
 465                 else if (vm_cpuset->which == VM_SUSPENDED_CPUS)
 466                         *cpuset = vm_suspended_cpus(sc->vm);
 467                 else
 468                         error = EINVAL;
 469                 if (error == 0)
 470                         error = copyout(cpuset, vm_cpuset->cpus, size);
 471                 free(cpuset, M_TEMP);
 472                 break;
 473         default:
 474                 error = ENOTTY;
 475                 break;
 476         }
 477
 478         if (state_changed == 1) {
 479                 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
 480         } else if (state_changed == 2) {
 481                 for (vcpu = 0; vcpu < VM_MAXCPU; vcpu++)
 482                         vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
 483         }
 484
 485 done:
 486         /* Make sure that no handler returns a bogus value like ERESTART */
 487         KASSERT(error >= 0, ("vmmdev_ioctl: invalid error return %d", error));
 488         return (error);
 489 }
 490
 491 static int
 492 vmmdev_mmap_single(struct cdev *cdev, vm_ooffset_t *offset,
 493                    vm_size_t size, struct vm_object **object, int nprot)
 494 {
 495         int error;
 496         struct vmmdev_softc *sc;
 497
 498         sc = vmmdev_lookup2(cdev);
 499         if (sc != NULL && (nprot & PROT_EXEC) == 0)
 500                 error = vm_get_memobj(sc->vm, *offset, size, offset, object);
 501         else
 502                 error = EINVAL;
 503
 504         return (error);
 505 }
 506
 507 static void
 508 vmmdev_destroy(void *arg)
 509 {
 510
 511         struct vmmdev_softc *sc = arg;
 512
 513         if (sc->cdev != NULL)
 514                 destroy_dev(sc->cdev);
 515
 516         if (sc->vm != NULL)
 517                 vm_destroy(sc->vm);
 518
 519         if ((sc->flags & VSC_LINKED) != 0) {
 520                 mtx_lock(&vmmdev_mtx);
 521                 SLIST_REMOVE(&head, sc, vmmdev_softc, link);
 522                 mtx_unlock(&vmmdev_mtx);
 523         }
 524
 525         free(sc, M_VMMDEV);
 526 }
 527
 528 static int
 529 sysctl_vmm_destroy(SYSCTL_HANDLER_ARGS)
 530 {
 531         int error;
 532         char buf[VM_MAX_NAMELEN];
 533         struct vmmdev_softc *sc;
 534         struct cdev *cdev;
 535
 536         strlcpy(buf, "beavis", sizeof(buf));
 537         error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
 538         if (error != 0 || req->newptr == NULL)
 539                 return (error);
 540
 541         mtx_lock(&vmmdev_mtx);
 542         sc = vmmdev_lookup(buf);
 543         if (sc == NULL || sc->cdev == NULL) {
 544                 mtx_unlock(&vmmdev_mtx);
 545                 return (EINVAL);
 546         }
 547
 548         /*
 549          * The 'cdev' will be destroyed asynchronously when 'si_threadcount'
 550          * goes down to 0 so we should not do it again in the callback.
 551          */
 552         cdev = sc->cdev;
 553         sc->cdev = NULL;
 554         mtx_unlock(&vmmdev_mtx);
 555
 556         /*
 557          * Schedule the 'cdev' to be destroyed:
 558          *
 559          * - any new operations on this 'cdev' will return an error (ENXIO).
 560          *
 561          * - when the 'si_threadcount' dwindles down to zero the 'cdev' will
 562          *   be destroyed and the callback will be invoked in a taskqueue
 563          *   context.
 564          */
 565         destroy_dev_sched_cb(cdev, vmmdev_destroy, sc);
 566
 567         return (0);
 568 }
 569 SYSCTL_PROC(_hw_vmm, OID_AUTO, destroy, CTLTYPE_STRING | CTLFLAG_RW,
 570             NULL, 0, sysctl_vmm_destroy, "A", NULL);
 571
 572 static struct cdevsw vmmdevsw = {
 573         .d_name         = "vmmdev",
 574         .d_version      = D_VERSION,
 575         .d_ioctl        = vmmdev_ioctl,
 576         .d_mmap_single  = vmmdev_mmap_single,
 577         .d_read         = vmmdev_rw,
 578         .d_write        = vmmdev_rw,
 579 };
 580
 581 static int
 582 sysctl_vmm_create(SYSCTL_HANDLER_ARGS)
 583 {
 584         int error;
 585         struct vm *vm;
 586         struct cdev *cdev;
 587         struct vmmdev_softc *sc, *sc2;
 588         char buf[VM_MAX_NAMELEN];
 589
 590         strlcpy(buf, "beavis", sizeof(buf));
 591         error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
 592         if (error != 0 || req->newptr == NULL)
 593                 return (error);
 594
 595         mtx_lock(&vmmdev_mtx);
 596         sc = vmmdev_lookup(buf);
 597         mtx_unlock(&vmmdev_mtx);
 598         if (sc != NULL)
 599                 return (EEXIST);
 600
 601         error = vm_create(buf, &vm);
 602         if (error != 0)
 603                 return (error);
 604
 605         sc = malloc(sizeof(struct vmmdev_softc), M_VMMDEV, M_WAITOK | M_ZERO);
 606         sc->vm = vm;
 607
 608         /*
 609          * Lookup the name again just in case somebody sneaked in when we
 610          * dropped the lock.
 611          */
 612         mtx_lock(&vmmdev_mtx);
 613         sc2 = vmmdev_lookup(buf);
 614         if (sc2 == NULL) {
 615                 SLIST_INSERT_HEAD(&head, sc, link);
 616                 sc->flags |= VSC_LINKED;
 617         }
 618         mtx_unlock(&vmmdev_mtx);
 619
 620         if (sc2 != NULL) {
 621                 vmmdev_destroy(sc);
 622                 return (EEXIST);
 623         }
 624
 625         error = make_dev_p(MAKEDEV_CHECKNAME, &cdev, &vmmdevsw, NULL,
 626                            UID_ROOT, GID_WHEEL, 0600, "vmm/%s", buf);
 627         if (error != 0) {
 628                 vmmdev_destroy(sc);
 629                 return (error);
 630         }
 631
 632         mtx_lock(&vmmdev_mtx);
 633         sc->cdev = cdev;
 634         sc->cdev->si_drv1 = sc;
 635         mtx_unlock(&vmmdev_mtx);
 636
 637         return (0);
 638 }
 639 SYSCTL_PROC(_hw_vmm, OID_AUTO, create, CTLTYPE_STRING | CTLFLAG_RW,
 640             NULL, 0, sysctl_vmm_create, "A", NULL);
 641
 642 void
 643 vmmdev_init(void)
 644 {
 645         mtx_init(&vmmdev_mtx, "vmm device mutex", NULL, MTX_DEF);
 646 }
 647
 648 int
 649 vmmdev_cleanup(void)
 650 {
 651         int error;
 652
 653         if (SLIST_EMPTY(&head))
 654                 error = 0;
 655         else
 656                 error = EBUSY;
 657
 658         return (error);
 659 }