2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/kernel.h>
34 #include <sys/queue.h>
36 #include <sys/mutex.h>
37 #include <sys/malloc.h>
39 #include <sys/sysctl.h>
40 #include <sys/libkern.h>
41 #include <sys/ioccom.h>
47 #include <vm/vm_map.h>
49 #include <machine/vmparam.h>
50 #include <machine/vmm.h>
51 #include <machine/vmm_instruction_emul.h>
52 #include <machine/vmm_dev.h>
54 #include "vmm_lapic.h"
58 #include "io/vatpic.h"
59 #include "io/vioapic.h"
64 struct vm *vm; /* vm instance cookie */
66 SLIST_ENTRY(vmmdev_softc) link;
69 #define VSC_LINKED 0x01
71 static SLIST_HEAD(, vmmdev_softc) head;
73 static struct mtx vmmdev_mtx;
75 static MALLOC_DEFINE(M_VMMDEV, "vmmdev", "vmmdev");
79 static struct vmmdev_softc *
80 vmmdev_lookup(const char *name)
82 struct vmmdev_softc *sc;
84 #ifdef notyet /* XXX kernel is not compiled with invariants */
85 mtx_assert(&vmmdev_mtx, MA_OWNED);
88 SLIST_FOREACH(sc, &head, link) {
89 if (strcmp(name, vm_name(sc->vm)) == 0)
96 static struct vmmdev_softc *
97 vmmdev_lookup2(struct cdev *cdev)
100 return (cdev->si_drv1);
104 vmmdev_rw(struct cdev *cdev, struct uio *uio, int flags)
106 int error, off, c, prot;
109 struct vmmdev_softc *sc;
111 static char zerobuf[PAGE_SIZE];
114 sc = vmmdev_lookup2(cdev);
118 prot = (uio->uio_rw == UIO_WRITE ? VM_PROT_WRITE : VM_PROT_READ);
119 while (uio->uio_resid > 0 && error == 0) {
120 gpa = uio->uio_offset;
121 off = gpa & PAGE_MASK;
122 c = min(uio->uio_resid, PAGE_SIZE - off);
125 * The VM has a hole in its physical memory map. If we want to
126 * use 'dd' to inspect memory beyond the hole we need to
127 * provide bogus data for memory that lies in the hole.
129 * Since this device does not support lseek(2), dd(1) will
130 * read(2) blocks of data to simulate the lseek(2).
132 hpa = vm_gpa_hold(sc->vm, gpa, c, prot, &cookie);
134 if (uio->uio_rw == UIO_READ)
135 error = uiomove(zerobuf, c, uio);
139 error = uiomove(hpa, c, uio);
140 vm_gpa_release(cookie);
147 vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
150 int error, vcpu, state_changed, size;
152 struct vmmdev_softc *sc;
153 struct vm_memory_segment *seg;
154 struct vm_register *vmreg;
155 struct vm_seg_desc *vmsegdesc;
156 struct vm_run *vmrun;
157 struct vm_exception *vmexc;
158 struct vm_lapic_irq *vmirq;
159 struct vm_lapic_msi *vmmsi;
160 struct vm_ioapic_irq *ioapic_irq;
161 struct vm_isa_irq *isa_irq;
162 struct vm_isa_irq_trigger *isa_irq_trigger;
163 struct vm_capability *vmcap;
164 struct vm_pptdev *pptdev;
165 struct vm_pptdev_mmio *pptmmio;
166 struct vm_pptdev_msi *pptmsi;
167 struct vm_pptdev_msix *pptmsix;
168 struct vm_nmi *vmnmi;
169 struct vm_stats *vmstats;
170 struct vm_stat_desc *statdesc;
171 struct vm_x2apic *x2apic;
172 struct vm_gpa_pte *gpapte;
173 struct vm_suspend *vmsuspend;
174 struct vm_gla2gpa *gg;
175 struct vm_activate_cpu *vac;
176 struct vm_cpuset *vm_cpuset;
177 struct vm_intinfo *vmii;
178 struct vm_rtc_time *rtctime;
179 struct vm_rtc_data *rtcdata;
181 sc = vmmdev_lookup2(cdev);
190 * Some VMM ioctls can operate only on vcpus that are not running.
194 case VM_GET_REGISTER:
195 case VM_SET_REGISTER:
196 case VM_GET_SEGMENT_DESCRIPTOR:
197 case VM_SET_SEGMENT_DESCRIPTOR:
198 case VM_INJECT_EXCEPTION:
199 case VM_GET_CAPABILITY:
200 case VM_SET_CAPABILITY:
203 case VM_SET_X2APIC_STATE:
205 case VM_ACTIVATE_CPU:
208 case VM_RESTART_INSTRUCTION:
210 * XXX fragile, handle with care
211 * Assumes that the first field of the ioctl data is the vcpu.
214 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
219 error = vcpu_set_state(sc->vm, vcpu, VCPU_FROZEN, true);
226 case VM_MAP_PPTDEV_MMIO:
228 case VM_UNBIND_PPTDEV:
232 * ioctls that operate on the entire virtual machine must
233 * prevent all vcpus from running.
236 for (vcpu = 0; vcpu < VM_MAXCPU; vcpu++) {
237 error = vcpu_set_state(sc->vm, vcpu, VCPU_FROZEN, true);
244 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
257 vmrun = (struct vm_run *)data;
258 error = vm_run(sc->vm, vmrun);
261 vmsuspend = (struct vm_suspend *)data;
262 error = vm_suspend(sc->vm, vmsuspend->how);
265 error = vm_reinit(sc->vm);
268 statdesc = (struct vm_stat_desc *)data;
269 error = vmm_stat_desc_copy(statdesc->index,
270 statdesc->desc, sizeof(statdesc->desc));
274 CTASSERT(MAX_VM_STATS >= MAX_VMM_STAT_ELEMS);
275 vmstats = (struct vm_stats *)data;
276 getmicrotime(&vmstats->tv);
277 error = vmm_stat_copy(sc->vm, vmstats->cpuid,
278 &vmstats->num_entries, vmstats->statbuf);
282 pptmsi = (struct vm_pptdev_msi *)data;
283 error = ppt_setup_msi(sc->vm, pptmsi->vcpu,
284 pptmsi->bus, pptmsi->slot, pptmsi->func,
285 pptmsi->addr, pptmsi->msg,
289 pptmsix = (struct vm_pptdev_msix *)data;
290 error = ppt_setup_msix(sc->vm, pptmsix->vcpu,
291 pptmsix->bus, pptmsix->slot,
292 pptmsix->func, pptmsix->idx,
293 pptmsix->addr, pptmsix->msg,
294 pptmsix->vector_control);
296 case VM_MAP_PPTDEV_MMIO:
297 pptmmio = (struct vm_pptdev_mmio *)data;
298 error = ppt_map_mmio(sc->vm, pptmmio->bus, pptmmio->slot,
299 pptmmio->func, pptmmio->gpa, pptmmio->len,
303 pptdev = (struct vm_pptdev *)data;
304 error = vm_assign_pptdev(sc->vm, pptdev->bus, pptdev->slot,
307 case VM_UNBIND_PPTDEV:
308 pptdev = (struct vm_pptdev *)data;
309 error = vm_unassign_pptdev(sc->vm, pptdev->bus, pptdev->slot,
312 case VM_INJECT_EXCEPTION:
313 vmexc = (struct vm_exception *)data;
314 error = vm_inject_exception(sc->vm, vmexc->cpuid,
315 vmexc->vector, vmexc->error_code_valid, vmexc->error_code,
316 vmexc->restart_instruction);
319 vmnmi = (struct vm_nmi *)data;
320 error = vm_inject_nmi(sc->vm, vmnmi->cpuid);
323 vmirq = (struct vm_lapic_irq *)data;
324 error = lapic_intr_edge(sc->vm, vmirq->cpuid, vmirq->vector);
326 case VM_LAPIC_LOCAL_IRQ:
327 vmirq = (struct vm_lapic_irq *)data;
328 error = lapic_set_local_intr(sc->vm, vmirq->cpuid,
332 vmmsi = (struct vm_lapic_msi *)data;
333 error = lapic_intr_msi(sc->vm, vmmsi->addr, vmmsi->msg);
335 case VM_IOAPIC_ASSERT_IRQ:
336 ioapic_irq = (struct vm_ioapic_irq *)data;
337 error = vioapic_assert_irq(sc->vm, ioapic_irq->irq);
339 case VM_IOAPIC_DEASSERT_IRQ:
340 ioapic_irq = (struct vm_ioapic_irq *)data;
341 error = vioapic_deassert_irq(sc->vm, ioapic_irq->irq);
343 case VM_IOAPIC_PULSE_IRQ:
344 ioapic_irq = (struct vm_ioapic_irq *)data;
345 error = vioapic_pulse_irq(sc->vm, ioapic_irq->irq);
347 case VM_IOAPIC_PINCOUNT:
348 *(int *)data = vioapic_pincount(sc->vm);
350 case VM_ISA_ASSERT_IRQ:
351 isa_irq = (struct vm_isa_irq *)data;
352 error = vatpic_assert_irq(sc->vm, isa_irq->atpic_irq);
353 if (error == 0 && isa_irq->ioapic_irq != -1)
354 error = vioapic_assert_irq(sc->vm,
355 isa_irq->ioapic_irq);
357 case VM_ISA_DEASSERT_IRQ:
358 isa_irq = (struct vm_isa_irq *)data;
359 error = vatpic_deassert_irq(sc->vm, isa_irq->atpic_irq);
360 if (error == 0 && isa_irq->ioapic_irq != -1)
361 error = vioapic_deassert_irq(sc->vm,
362 isa_irq->ioapic_irq);
364 case VM_ISA_PULSE_IRQ:
365 isa_irq = (struct vm_isa_irq *)data;
366 error = vatpic_pulse_irq(sc->vm, isa_irq->atpic_irq);
367 if (error == 0 && isa_irq->ioapic_irq != -1)
368 error = vioapic_pulse_irq(sc->vm, isa_irq->ioapic_irq);
370 case VM_ISA_SET_IRQ_TRIGGER:
371 isa_irq_trigger = (struct vm_isa_irq_trigger *)data;
372 error = vatpic_set_irq_trigger(sc->vm,
373 isa_irq_trigger->atpic_irq, isa_irq_trigger->trigger);
376 seg = (struct vm_memory_segment *)data;
377 error = vm_malloc(sc->vm, seg->gpa, seg->len);
379 case VM_GET_MEMORY_SEG:
380 seg = (struct vm_memory_segment *)data;
382 (void)vm_gpabase2memseg(sc->vm, seg->gpa, seg);
385 case VM_GET_REGISTER:
386 vmreg = (struct vm_register *)data;
387 error = vm_get_register(sc->vm, vmreg->cpuid, vmreg->regnum,
390 case VM_SET_REGISTER:
391 vmreg = (struct vm_register *)data;
392 error = vm_set_register(sc->vm, vmreg->cpuid, vmreg->regnum,
395 case VM_SET_SEGMENT_DESCRIPTOR:
396 vmsegdesc = (struct vm_seg_desc *)data;
397 error = vm_set_seg_desc(sc->vm, vmsegdesc->cpuid,
401 case VM_GET_SEGMENT_DESCRIPTOR:
402 vmsegdesc = (struct vm_seg_desc *)data;
403 error = vm_get_seg_desc(sc->vm, vmsegdesc->cpuid,
407 case VM_GET_CAPABILITY:
408 vmcap = (struct vm_capability *)data;
409 error = vm_get_capability(sc->vm, vmcap->cpuid,
413 case VM_SET_CAPABILITY:
414 vmcap = (struct vm_capability *)data;
415 error = vm_set_capability(sc->vm, vmcap->cpuid,
419 case VM_SET_X2APIC_STATE:
420 x2apic = (struct vm_x2apic *)data;
421 error = vm_set_x2apic_state(sc->vm,
422 x2apic->cpuid, x2apic->state);
424 case VM_GET_X2APIC_STATE:
425 x2apic = (struct vm_x2apic *)data;
426 error = vm_get_x2apic_state(sc->vm,
427 x2apic->cpuid, &x2apic->state);
429 case VM_GET_GPA_PMAP:
430 gpapte = (struct vm_gpa_pte *)data;
431 pmap_get_mapping(vmspace_pmap(vm_get_vmspace(sc->vm)),
432 gpapte->gpa, gpapte->pte, &gpapte->ptenum);
435 case VM_GET_HPET_CAPABILITIES:
436 error = vhpet_getcap((struct vm_hpet_cap *)data);
439 CTASSERT(PROT_READ == VM_PROT_READ);
440 CTASSERT(PROT_WRITE == VM_PROT_WRITE);
441 CTASSERT(PROT_EXEC == VM_PROT_EXECUTE);
442 gg = (struct vm_gla2gpa *)data;
443 error = vm_gla2gpa(sc->vm, gg->vcpuid, &gg->paging, gg->gla,
445 KASSERT(error == 0 || error == 1 || error == -1,
446 ("%s: vm_gla2gpa unknown error %d", __func__, error));
449 * error = 0: the translation was successful
450 * error = 1: a fault was injected into the guest
459 case VM_ACTIVATE_CPU:
460 vac = (struct vm_activate_cpu *)data;
461 error = vm_activate_cpu(sc->vm, vac->vcpuid);
465 vm_cpuset = (struct vm_cpuset *)data;
466 size = vm_cpuset->cpusetsize;
467 if (size < sizeof(cpuset_t) || size > CPU_MAXSIZE / NBBY) {
471 cpuset = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
472 if (vm_cpuset->which == VM_ACTIVE_CPUS)
473 *cpuset = vm_active_cpus(sc->vm);
474 else if (vm_cpuset->which == VM_SUSPENDED_CPUS)
475 *cpuset = vm_suspended_cpus(sc->vm);
479 error = copyout(cpuset, vm_cpuset->cpus, size);
480 free(cpuset, M_TEMP);
483 vmii = (struct vm_intinfo *)data;
484 error = vm_exit_intinfo(sc->vm, vmii->vcpuid, vmii->info1);
487 vmii = (struct vm_intinfo *)data;
488 error = vm_get_intinfo(sc->vm, vmii->vcpuid, &vmii->info1,
492 rtcdata = (struct vm_rtc_data *)data;
493 error = vrtc_nvram_write(sc->vm, rtcdata->offset,
497 rtcdata = (struct vm_rtc_data *)data;
498 error = vrtc_nvram_read(sc->vm, rtcdata->offset,
502 rtctime = (struct vm_rtc_time *)data;
503 error = vrtc_set_time(sc->vm, rtctime->secs);
507 rtctime = (struct vm_rtc_time *)data;
508 rtctime->secs = vrtc_get_time(sc->vm);
510 case VM_RESTART_INSTRUCTION:
511 error = vm_restart_instruction(sc->vm, vcpu);
518 if (state_changed == 1) {
519 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
520 } else if (state_changed == 2) {
521 for (vcpu = 0; vcpu < VM_MAXCPU; vcpu++)
522 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
526 /* Make sure that no handler returns a bogus value like ERESTART */
527 KASSERT(error >= 0, ("vmmdev_ioctl: invalid error return %d", error));
532 vmmdev_mmap_single(struct cdev *cdev, vm_ooffset_t *offset,
533 vm_size_t size, struct vm_object **object, int nprot)
536 struct vmmdev_softc *sc;
538 sc = vmmdev_lookup2(cdev);
539 if (sc != NULL && (nprot & PROT_EXEC) == 0)
540 error = vm_get_memobj(sc->vm, *offset, size, offset, object);
548 vmmdev_destroy(void *arg)
551 struct vmmdev_softc *sc = arg;
553 if (sc->cdev != NULL)
554 destroy_dev(sc->cdev);
559 if ((sc->flags & VSC_LINKED) != 0) {
560 mtx_lock(&vmmdev_mtx);
561 SLIST_REMOVE(&head, sc, vmmdev_softc, link);
562 mtx_unlock(&vmmdev_mtx);
569 sysctl_vmm_destroy(SYSCTL_HANDLER_ARGS)
572 char buf[VM_MAX_NAMELEN];
573 struct vmmdev_softc *sc;
576 strlcpy(buf, "beavis", sizeof(buf));
577 error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
578 if (error != 0 || req->newptr == NULL)
581 mtx_lock(&vmmdev_mtx);
582 sc = vmmdev_lookup(buf);
583 if (sc == NULL || sc->cdev == NULL) {
584 mtx_unlock(&vmmdev_mtx);
589 * The 'cdev' will be destroyed asynchronously when 'si_threadcount'
590 * goes down to 0 so we should not do it again in the callback.
594 mtx_unlock(&vmmdev_mtx);
597 * Schedule the 'cdev' to be destroyed:
599 * - any new operations on this 'cdev' will return an error (ENXIO).
601 * - when the 'si_threadcount' dwindles down to zero the 'cdev' will
602 * be destroyed and the callback will be invoked in a taskqueue
605 destroy_dev_sched_cb(cdev, vmmdev_destroy, sc);
609 SYSCTL_PROC(_hw_vmm, OID_AUTO, destroy, CTLTYPE_STRING | CTLFLAG_RW,
610 NULL, 0, sysctl_vmm_destroy, "A", NULL);
612 static struct cdevsw vmmdevsw = {
614 .d_version = D_VERSION,
615 .d_ioctl = vmmdev_ioctl,
616 .d_mmap_single = vmmdev_mmap_single,
618 .d_write = vmmdev_rw,
622 sysctl_vmm_create(SYSCTL_HANDLER_ARGS)
627 struct vmmdev_softc *sc, *sc2;
628 char buf[VM_MAX_NAMELEN];
630 strlcpy(buf, "beavis", sizeof(buf));
631 error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
632 if (error != 0 || req->newptr == NULL)
635 mtx_lock(&vmmdev_mtx);
636 sc = vmmdev_lookup(buf);
637 mtx_unlock(&vmmdev_mtx);
641 error = vm_create(buf, &vm);
645 sc = malloc(sizeof(struct vmmdev_softc), M_VMMDEV, M_WAITOK | M_ZERO);
649 * Lookup the name again just in case somebody sneaked in when we
652 mtx_lock(&vmmdev_mtx);
653 sc2 = vmmdev_lookup(buf);
655 SLIST_INSERT_HEAD(&head, sc, link);
656 sc->flags |= VSC_LINKED;
658 mtx_unlock(&vmmdev_mtx);
665 error = make_dev_p(MAKEDEV_CHECKNAME, &cdev, &vmmdevsw, NULL,
666 UID_ROOT, GID_WHEEL, 0600, "vmm/%s", buf);
672 mtx_lock(&vmmdev_mtx);
674 sc->cdev->si_drv1 = sc;
675 mtx_unlock(&vmmdev_mtx);
679 SYSCTL_PROC(_hw_vmm, OID_AUTO, create, CTLTYPE_STRING | CTLFLAG_RW,
680 NULL, 0, sysctl_vmm_create, "A", NULL);
685 mtx_init(&vmmdev_mtx, "vmm device mutex", NULL, MTX_DEF);
693 if (SLIST_EMPTY(&head))